How to use it
Filter by country, focus area or way to engage to find groups near your problem.
Open an entry for its leading researchers, flagship outputs, funding and how to work with it: PhDs, partnerships, grants.
Researchers marked Top 2% are verified on the 2026 Stanford–Elsevier list of the most-cited scientists.
76groups, institutes and funders
19countries
107researchers on the Stanford–Elsevier top-2% list
460sources cited
3IA Côte d'Azur (IA Cluster)
PhD and postdoc candidates (three funded calls a year via chairholders), SMEs and large firms seeking co-run industrial chairs, and medical-AI teams wanting links with Inria/Inserm groups in Sophia Antipolis.
Medium confidenceDetailsClose
One of France's four original 3IA institutes (2019), re-labelled as the '3IA Côte d'Azur 2030' IA Cluster in 2024; 38 academic chairs and fellows across core AI, AI for health and AI for secure and smart spaces.
Strengths
- Regular, transparent funded-position calls with stated salaries and durations
- Strong AI-for-health line building on Inria and Inserm medical-imaging groups
- Industry traction: 116 signed contracts, €16.8m invested by companies, 12 startups
- Experienced new leadership (Nov 2025) drawn from the institute's own chairholders and executive team
Watch-outs
- Modest state award (€20m to 2030) compared with Paris and Grenoble clusters
- PhD/postdoc calls are only open through existing chairholders, not to open applications
- Smaller ecosystem than Paris; EFELIA training school split off as a separate entity in Jan 2025
Leading researchers
- Director, 3IA Côte d'Azur (since 4 Nov 2025)Succeeded Charles Bouveyron; Executive Director of 3IA since 2019
- Scientific Director, 3IA Côte d'Azur (since Nov 2025); CNRS research directorHeads Inria's MARIANNE team (argumentation, NLP); 3IA chairholder since 2019; succeeded Nicholas Ayache
- Website
- 3ia.univ-cotedazur.eu
- Kind
- Network / society
- Institution
- Université Côte d'Azur (coordinator) with CNRS, Inria, Inserm, EURECOM and SKEMA Business School
- Location
- Nice / Sophia Antipolis, France
- Founded
- 2019
- Funding scale
- €20m France 2030 IA Cluster grant for 2024–2030, after a €19m 3IA grant (2019–2023); regional firms forecast ~€75m over 7 years (May 2024 forecast, not committed); €16.8m company investment to date (homepage)
Flagship outputs
- Funded PhD (36 months) and postdoc (24 months) calls: 3 rounds in 2026, 16 PhD and 22 postdoc slots
- Industrial chairs: up to 4-year joint programmes co-run by a 3IA chairholder and a company researcher
- 2025 call for Senior and Junior Chairs (4-year terms, budget secured to 2030; applications 6 Jan–28 Feb 2025, closed)
Focus areas
- ML for science
- Probabilistic ML
- NLP / LLMs
- Trustworthy / safe AI
- Computer vision
How to engage
- PhD / postdoc calls via a 3IA chairholder (three rounds a year; 2026 Call 3 deadline was 18 Sep 2026)
- Industrial chairs (SMEs, mid-caps and large firms with R&D)
- Chair and invited-researcher calls
- Data-scientist engineer support (3IA Techpool)
Sources · verified November 2026
Anchor: #eu-3ia-cote-dazur
Amsterdam Machine Learning Lab (AMLab) & ELLIS Unit Amsterdam
Companies willing to co-fund an industry lab (ICAI/QUvA model), and teams wanting generative, geometric or probabilistic-ML expertise or AI-for-materials collaborators.
High confidenceDetailsClose
UvA's AMLab is known for generative models, geometric deep learning and probabilistic inference; the ELLIS Unit Amsterdam (UvA, VU, CWI; est. 2019) and the ICAI industry-lab network extend it.
Strengths
- Origin of foundational generative and graph-learning methods (VAE, GCN, group-equivariant CNNs)
- Proven industry-lab model: QUvA (Qualcomm), Delta Lab (Bosch), AI4Science Lab, and 60 ICAI labs with 160 partners nationwide
- Strong deep-tech spin-out signal (CuspAI, co-founded by Max Welling)
- Balanced CSRankings profile: top-12 in ML core with top-10 NLP and AI
Watch-outs
- AMLab itself is small (6 listed faculty); reputation leans on a few senior PIs
- Industry-lab model favours partners able to co-fund multi-year PhD cohorts
- ML-core output per faculty is below Swiss/UK peers in CSRankings
Leading researchers
- Max WellingTop 2%Professor, AMLab faculty; co-founder of CuspAIin the V9 (Aug 2026) list per the topcitedscientists.com mirror (UvA; AI & Image Processing; NL ICT #1). VAEs, graph conv nets; ICLR 2024 test-of-time award; co-founder of CuspAI
- Director, AMLabProbabilistic programming
- Cees SnoekTop 2%Professor, UvA Informatics Institute (not AMLab faculty)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (UvA; NL ICT #51). Video and multimodal vision; co-founded ICAI
- Faculty, AMLabGeometric / equivariant deep learning
- Professor, UvA Informatics Institute (not AMLab faculty)Information retrieval and recommender systems
- Website
- amlab.science.uva.nl
- Kind
- University group / lab
- Institution
- University of Amsterdam
- Location
- Amsterdam, Netherlands
- CSRankings signal
- #12 in Europe for ML core (30.3 adj.), #9 for NLP, #8 for AI; #12 combined, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Probabilistic ML
- ML for science
- Computer vision
- RL
- Causality
- NLP / LLMs
How to engage
- PhD / ELLIS PhD programme
- ICAI co-funded industry lab
- Spin-out / tech transfer
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-uva-amlab-ellis-amsterdam
ANITI – Artificial and Natural Intelligence Toulouse Institute (IA Cluster)
Teams working on certifiable or safety-critical ML (aerospace, energy, space), companies wanting industrial chairs with Toulouse labs, and researchers seeking trustworthy-AI chairs or PhD/postdoc posts.
Medium confidenceDetailsClose
Toulouse's France 2030 IA Cluster (labelled May 2024, launched 7 Jul 2025) succeeding the 2019 3IA institute; focused on trustworthy, certifiable and embeddable AI for aerospace, energy and environment; ~300 researchers and 19 chairs.
Strengths
- Clear specialism: trustworthy, certifiable and embeddable AI with real aerospace and energy use cases
- Strong industry anchor: Airbus, Thales entities and Vitesco are in the ANR consortium; Safran, RTE and CNES are named at launch; ANITI reports 60+ partners
- Mature open-source portfolio for robustness, explainability and conformal prediction (deel-ai)
- Concrete chair structure: 19 chairs incl. industrial and international attractiveness chairs
Watch-outs
- Smaller IA Cluster award (€20m) than Paris or Grenoble; per-chair budgets not published
- Leadership is poorly documented online: the governance page names only the steering chair, and the scientific director's current title is not shown
- Applied, regional-industry orientation; less presence in frontier-model or core-ML venues
Leading researchers
- Coordinator of the ANITI IA Cluster grant (ANR, Dec 2024–Dec 2029); ANITI scientific director since 1 Apr 2023 (succeeding Nicholas Asher) per the 2023 announcement; current title not shown on ANITI's governance or launch pagesApplied mathematician (INP Toulouse); ANR lists him as coordinator of the €20m cluster award
- Chair, ANITI Steering CommitteeChairs the institute's decision-making body
- International Chair holder (RL4Robots)Reinforcement learning for robotics; one of 19 cluster chairs
- Individual Chair holder (MoAI – Center for Moral AI)Moral psychology of machines; one of 19 cluster chairs
- Website
- aniti.univ-toulouse.fr
- Kind
- Network / society
- Institution
- Université de Toulouse (ComUE, coordinator) with 16 academic labs incl. CNRS, INRAE, ONERA and IRT Saint Exupéry
- Location
- Toulouse, France
- Founded
- 2019
- Funding scale
- €20m France 2030 IA Cluster grant (ANR, Dec 2024–Dec 2029); ~€90m total contributions cited by ANITI; earlier 3IA grant €22m (2019–2026)
Flagship outputs
- Xplique: MIT-licensed neural-network explainability toolbox from the DEEL team, ANITI-funded (~750 GitHub stars)
- ANITI open-source library list: deel-lip, PUNCC, OODEEL, Influenciae, toulbar2, Pinocchio and more
- 19 IA Cluster chairs (individual, synergy, industrial, international) in trustworthy AI and AI+X
Focus areas
- Trustworthy / safe AI
- Efficient ML
- ML for science
- RL
- Theory
How to engage
- Industrial chairs (6 of the 19 cluster chairs, e.g., certification of embedded ML, Earth observation)
- Research chairs via 'Choose France for Science' with Inria (contact the scientific director)
- PhD, postdoc and internship offers
- Open-source libraries (deel-ai, toulbar2 and others)
- AI for Health M2 internship calls (Occitanie Region; 4th call Oct 2026)
Sources · verified November 2026
Anchor: #eu-aniti-toulouse
Archimedes – AI, Data Science and Algorithms Unit (Athena Research Center)
Researchers in theory, algorithms, learning theory and NLP looking for Greek collaborators; PhD candidates and visiting researchers; diaspora PIs wanting a Greece-based project.
Medium confidenceDetailsClose
Research unit of Greece's Athena Research Center (est. Jan 2022) for AI, data science and algorithms, co-founded by Timos Sellis, Christos Papadimitriou and Constantinos Daskalakis and funded by ~€19–21m from the EU Recovery and Resilience Facility.
Strengths
- World-class co-founders in algorithms and learning theory (Papadimitriou, Daskalakis)
- Funds PhDs (25 doctoral scholarships reported by Oct 2023) and visiting researchers, building a Greek AI talent pipeline
- Active in 2026: Athena RC signed an MoU with Greece's Special Secretariat for AI and Data Governance (18 May 2026) in which Archimedes has an implementation role, and Archimedes/IMSI researchers won a BeyondFacts '26 best paper award
Watch-outs
- Start-up funding from the RRF covered only the first four years; long-term funding not stated on public pages
- Public site lists no current leadership, headcount or research groups
- Theory-leaning profile; less visible in large-scale model building or industry partnerships
Leading researchers
- Co-founder; Executive Board member; named head of the unit at its 2022 launchAthena Research Center; named head of the Unit in 2022
- Co-founder; Executive Board memberDonovan Family Professor of Computer Science, Columbia University; member of the Academy of Athens (reception May 2025)
- Co-founder and chief scientist; Executive Board memberAvanessians Professor of EECS, MIT (CSAIL)
- Website
- archimedesai.gr
- Kind
- Research institute
- Institution
- Athena Research Center
- Location
- Athens, Greece
- Founded
- 2022
- Funding scale
- ~€19m (Greek News Agenda, 2022) to over €21m (Hellenic AI Society, 2022) from the EU Recovery and Resilience Facility / Greece 2.0, covering the first four years
Flagship outputs
Focus areas
- Theory
- NLP / LLMs
- Computer vision
- ML for science
How to engage
- PhD funding
- Visiting researcher
- Research collaboration (lead-PI projects)
- Events and summer schools
Sources · verified November 2026
Anchor: #eu-archimedes-athena-rc
ARIA – Advanced Research and Invention Agency (AI-relevant programmes)
Researchers, startups and companies working on AI assurance/formal verification, agent security, energy-efficient AI hardware or robotics who want large, milestone-driven grants.
High confidenceDetailsClose
UK high-risk research funder (est. Jan 2023, ≥£1bn for 2025–29) whose AI-relevant programmes include Safeguarded AI (£59m, formal-methods assurance), Scaling Trust (~£50m, secure AI agents), Scaling Compute and robotics.
Strengths
- Rare funder of AI assurance and formal verification at scale (£59m Safeguarded AI)
- Applicant pool is broad: individuals, startups, companies, universities and charities
- Rolling calls (Scaling Trust reviewed in quarterly batches) that accept individuals, companies, universities and charities
Watch-outs
- Programmes pivot quickly: Safeguarded AI cancelled its TA3 Phase 2 call (Feb 2026) and changed director (Apr 2026)
- Exempt from Freedom of Information requests; 2026 press criticism of grants to US firms and VCs
- AI is one of several opportunity spaces; ARIA expects to fund few projects per cycle and may award none
Leading researchers
- CEO (since Feb 2026)Succeeded founding CEO Ilan Gur
- Programme Director, Safeguarded AI (from Apr 2026)Took over from David 'davidad' Dalrymple, now Technical Advisor
- Programme Director, Scaling TrustListed as Scaling Trust programme director on ARIA's homepage
- Website
- aria.org.uk
- Kind
- Foundation / funder
- Institution
- UK non-departmental public body sponsored by DSIT
- Location
- London, United Kingdom
- Founded
- 2023
- Funding scale
- Initial ~£800m; at least £1bn over the 2025–29 Spending Review; Safeguarded AI £59m; Scaling Trust nearly £50m
Flagship outputs
Focus areas
- Trustworthy / safe AI
- Efficient ML
- Robotics learning
- Theory
How to engage
- Grants / calls (rolling, reviewed quarterly)
- Open to individuals, universities, companies and charities
- Programme Director recruitment
Sources · verified November 2026
Anchor: #eu-aria-uk
BCAM – Basque Center for Applied Mathematics (Data Science & AI area, Machine Learning line)
Companies wanting mathematically grounded ML (robust classification, time series, optimisation) via joint labs or industrial PhDs, and PhD/postdoc candidates in ML theory and probabilistic ML in Spain.
Medium confidenceDetailsClose
Severo Ochoa-accredited applied-mathematics institute whose Data Science & AI area includes a Machine Learning line (robust/minimax learning, probabilistic ML, concept drift), led by scientific director Jose A. Lozano.
Strengths
- Severo Ochoa Centre of Excellence (third consecutive accreditation, 2023–2027)
- Distinctive ML-theory profile: minimax/robust classification, concept drift, weak supervision
- Proven industrial PhD track with BBVA AI Factory, Iberdrola and BMW Group
- Training pipeline: 38 ongoing PhD theses, 27 postdocs, 8 MSCA and 4 ERC grants (2025)
Watch-outs
- Small: ML is one line within a 137-person maths institute (2025); total funding ~€7.2m in 2025
- Little private funding (€0.29m in 2025) and no current industrial PhDs listed
- Low visibility at top ML venues; open-source tools have small user bases (MRCpy ~30 GitHub stars)
Leading researchers
- Scientific Director, BCAM; coordinator, Data Science & AI area; ML line leadCo-leads the Severo Ochoa Strategic Lab on Trustworthy ML with Novi Quadrianto (Sussex)
- Ikerbasque Research Professor, BCAM Machine Learning lineMinimax risk classifiers (MRCpy); SEIO–BBVA Foundation best-contribution awards 2022 and 2025
- Ikerbasque Research Fellow, BCAM Machine Learning groupProbabilistic ML, bandits, Gaussian processes and clinical AI; formerly Columbia Data Science Institute
- Website
- bcamath.org
- Kind
- Research institute
- Institution
- Basque Excellence Research Centre (BERC) founded via Ikerbasque, Basque Government, UPV/EHU, Bizkaia and Innobasque
- Location
- Bilbao, Spain
- Founded
- 2008
- Funding scale
- €7.18m total funding in 2025 (€3.39m Basque, €2.8m Spain, €0.7m Europe, €0.29m private); Severo Ochoa accreditation 2023–2027
Flagship outputs
Focus areas
- Theory
- Probabilistic ML
- Trustworthy / safe AI
- ML for science
How to engage
- Industrial PhDs (past partners: BBVA AI Factory, Iberdrola, BMW Group, ITP Aero)
- Joint research lab on AI with Tecnalia and UPV/EHU; joint positions
- PhD, postdoc, master's-thesis and internship offers (BCAM Talent Hub)
- Knowledge Transfer Unit (contract research)
Sources · verified November 2026
Anchor: #eu-bcam-machine-learning
BIFOLD – Berlin Institute for the Foundations of Learning and Data
Partners needing explainable AI, ML for quantum chemistry/physics, earth-observation ML or scalable data systems for ML; recruiters targeting Berlin PhDs.
Medium confidenceDetailsClose
Permanently funded (€22m/yr since 2022) federal/Berlin centre at TU Berlin, founded 2020, for foundational research in machine learning and big-data management; co-directed by Klaus-Robert Müller and Volker Markl; 772 publications since 2020.
Strengths
- Permanent institutional funding of €22m/yr from the federal government and Berlin (since July 2022)
- Internationally recognised explainable-AI school (Müller, Samek, Montavon)
- Unusual pairing of ML with database/data-systems research (Markl)
- Broad Berlin consortium: TU, HU, FU, Charité, Fraunhofer HHI, DFKI, HPI, MDC
- Output volume: 772 publications and 34 dissertations reported since 2020
Watch-outs
- Mid-table ML-core output (TU Berlin #25 in Europe) versus South-German peers
- Many partner institutions; contracting partner depends on PI affiliation
- Contracting counterparty varies (TU Berlin, HU, Fraunhofer HHI, DFKI, ZIB) depending on the PI
Leading researchers
- Klaus-Robert MüllerTop 2%Co-director, BIFOLD; ELLIS Unit Berlin co-director; 2026 Leibniz Prizein the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under BIFOLD; Germany ICT #8). Kernel methods, explainable AI and ML for quantum chemistry
- Co-director, BIFOLDData management systems for ML at scale
- Wojciech SamekTop 2%ELLIS Unit Berlin co-director (Fraunhofer HHI & TU Berlin)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (Fraunhofer HHI; AI & Image Processing; global rank #4,793). Explainable AI (relevance propagation)
- ELLIS Unit Berlin co-director (TU Berlin)ML for remote sensing / earth observation
- Professor, TU BerlinOptimisation for ML; top TU Berlin ML-core output in CSRankings
- Website
- bifold.berlin
- Kind
- Research institute
- Institution
- TU Berlin (with HU Berlin, DFKI, HPI, MDC, Max Planck institutes)
- Location
- Berlin, Germany
- Founded
- 2020
- CSRankings signal
- TU Berlin: #25 in Europe for ML core (15.8 adj.), 2021–2025 (CSRankings data); Fraunhofer HHI/FU Berlin partners counted separately
- Funding scale
- €22m per year permanent institutional funding (federal government and Berlin, 50/50) since 1 July 2022
Flagship outputs
Focus areas
- Trustworthy / safe AI
- ML for science
- Efficient ML
- Computer vision
How to engage
- PhD / ELLIS PhD programme
- Industry partnership programme
- Spin-out / tech transfer
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-bifold-berlin
Bocconi University Department of Computing Sciences
Researchers and firms interested in the theory of neural networks, NLP/computational social science and game-theoretic learning, plus AI MSc graduates; finance/economics applications fit the business-school setting.
Medium confidenceDetailsClose
Young department combining computer science, mathematics and physics, known for statistical-physics approaches to learning and NLP; Bocconi is a partner in ELLIS Unit Milan (est. Oct 2021).
Strengths
- Distinctive statistical-physics school of learning theory (Zecchina, Mézard)
- Strong NLP/computational-social-science group (#19 NLP in Europe, CSRankings)
- Part of ELLIS Unit Milan alongside Politecnico di Milano, UniMi and Milano-Bicocca
- Business-school setting eases links to finance and economics applications
Watch-outs
- Small department; limited applied vision/robotics capacity
- Mid-table ML-core output (#26 in Europe) relative to theory reputation
- Department founding date and research funding not published on its site; head of department is Alon Rosen
Leading researchers
- Full Professor; ELLIS Unit Milan co-directorStatistical physics of learning and optimisation
- Marc MézardTop 2%Full Professorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Bocconi; subfield Fluids & Plasmas). Statistical physics, spin glasses and learning theory
- Dirk HovyTop 2%Associate Professorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Bocconi; AI & Image Processing; Italy ICT #11). NLP and computational social science; top Bocconi NLP output
- Assistant Professor; ELLIS ScholarOnline learning and algorithmic game theory
- Website
- cs.unibocconi.eu
- Kind
- University group / lab
- Institution
- Bocconi University
- Location
- Milan, Italy
- CSRankings signal
- #26 in Europe for ML core (14.9 adj.), #19 for NLP, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Theory
- NLP / LLMs
- RL
- ML for science
How to engage
- PhD / ELLIS PhD programme
- Industry partnership programme
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-bocconi-computing-sciences
CAIRNE (formerly CLAIRE) – Confederation of Laboratories for AI Research in Europe
Organisations wanting broad European AI-community reach, Horizon Europe consortium partners and policy engagement rather than a single lab.
Medium confidenceDetailsClose
Bottom-up, human-centred European AI research network launched as CLAIRE in 2018 and renamed CAIRNE after a trademark challenge; an AISBL with 500+ member groups and offices in nine cities.
Strengths
- Breadth: 500+ member groups and 260+ early-career researchers in its networks
- Formal support from nine national governments (e.g., Netherlands, Italy, Spain, Finland)
- Offices across Europe (The Hague HQ, Brussels, Paris, Rome, Zürich and more)
- Active in EU projects and policy events (e.g., Apply AI Summit)
Watch-outs
- Network and advocacy body; does not itself fund or run large research programmes
- Recent rename (CLAIRE → CAIRNE) may cause confusion in older references
- Budget and leadership not disclosed on its homepage
- Website
- cairne.eu
- Kind
- Network / society
- Location
- The Hague, Netherlands
- Founded
- 2018
Flagship outputs
Focus areas
- Trustworthy / safe AI
- Robotics learning
- NLP / LLMs
How to engage
- Membership (research & innovation networks)
- EU Horizon consortium partnering
- Events and policy workshops
Sources · verified November 2026
Anchor: #eu-cairne-claire
Cambridge Machine Learning (MLG / ELLIS Unit Cambridge)
Healthcare-AI, probabilistic-modelling and molecular-ML companies seeking PI collaboration or PhD/postdoc hires; investors scouting Bayesian-ML spin-outs; engage per PI (MLG, Computer Lab or van der Schaar Lab).
Medium confidenceDetailsClose
Cambridge's ML community spans the Engineering Department's Machine Learning Group (probabilistic ML) and the Computer Lab. The ELLIS Unit Cambridge centres on Bayesian and probabilistic ML.
Strengths
- Long-standing global centre for Bayesian/probabilistic ML (Gaussian processes, probabilistic programming)
- #2 ML-core output in Europe and #3 NLP output (CSRankings, 2021–2025)
- ELLIS Unit with stated application focus on language modelling, healthcare, systems and molecular modelling
- Dense local ecosystem (Microsoft Research Cambridge and DeepMind researchers interact with the group)
Watch-outs
- Output is concentrated: one PI accounts for ~38% of Cambridge's ML-core CSRankings count
- ML is spread across Engineering, Computer Lab and other departments; no single front door
- No founding date or single institutional front door for ML: partners must contract with the specific department (Engineering, Computer Lab or DAMTP) hosting the PI
Leading researchers
- Zoubin GhahramaniTop 2%Professor, MLGin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Cambridge; AI & Image Processing; UK ICT #6). Founding figure in Bayesian ML; listed MLG faculty
- Carl Edward RasmussenTop 2%Professor, MLGin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Cambridge; AI & Image Processing). Gaussian processes; co-author of the standard GP textbook
- Professor (outside MLG; leads the van der Schaar Lab)ML for medicine; highest ML-core output of any European academic in CSRankings 2021-25
- FacultyProbabilistic ML and molecular design
- Faculty, Department of Computer Science and Technology (Computer Lab)Gaussian processes, data science policy
- Website
- mlg.eng.cam.ac.uk
- Kind
- University group / lab
- Institution
- University of Cambridge
- Location
- Cambridge, United Kingdom
- CSRankings signal
- #2 in Europe for ML core (ICML/NeurIPS/ICLR), 2021–2025: 91.8 adjusted pubs, 31 faculty; #3 for NLP (CSRankings data)
Flagship outputs
Focus areas
- Probabilistic ML
- ML for science
- NLP / LLMs
- Trustworthy / safe AI
How to engage
- PhD / ELLIS PhD programme
- Industry partnership programme
- Spin-out / tech transfer
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-cambridge-machine-learning
Carl Zeiss Foundation (Carl-Zeiss-Stiftung, CZS)
Professors at universities and universities of applied sciences in Baden-Württemberg, Rhineland-Palatinate and Thuringia seeking multi-million AI research grants; partners of those groups.
Medium confidenceDetailsClose
Foundation established by Ernst Abbe in 1889 and sole owner of Carl Zeiss AG and Schott AG; AI is a priority topic with €140.9m invested across 56 ongoing projects, funded via competitive calls to universities in three states.
Strengths
- Large dedicated AI portfolio: €140.9m across 56 ongoing projects
- Large single grants (up to €5m, 5–6 years) suited to building new AI centres, e.g., CZS Heidelberg Center for Model-Based AI
- Covers the full spectrum from foundations and hardware to robotics and domain applications
- Senior AI scientific advisory board (André, Burgard, Ferrein, Schnabel, Schulz; met May 2026) shapes future calls
Watch-outs
- Restricted to institutions in Baden-Württemberg, Rhineland-Palatinate and Thuringia; quotas per university
- No direct route for companies or researchers abroad; access only as partners of eligible universities
- Calls are periodic and topic-bound, with long lead times
Leading researchers
- Scientific Advisory Board member (AI focus)Named SAB member at the May 2026 AI-focus meeting; informs design of future AI calls
- Scientific Advisory Board member (AI focus)Named Scientific Advisory Board member at the May 2026 meeting devoted to the AI focus topic
- Scientific Advisory Board member (AI focus)Named Scientific Advisory Board member at the May 2026 meeting devoted to the AI focus topic
- Website
- carl-zeiss-stiftung.de
- Kind
- Foundation / funder
- Location
- Stuttgart (admin); seats in Heidenheim an der Brenz and Jena, Germany
- Founded
- 1889
- Funding scale
- €140.9m invested across 56 ongoing AI projects (CZS AI topic page, viewed Nov 2026; page undated); single grants up to €5m
Flagship outputs
- CZS Durchbrüche (Breakthroughs) AI calls: up to €5m per project over 5–6 years (2025 round 'AI and environment' closed Apr 2025; topic-bound, periodic)
- CZS Focus 2027: Domains – applied-AI call for universities of applied sciences (€1.2m; full proposals due 5 Nov 2026; projects start May–Sep 2027)
- €5m for safe AI in healthcare at Tübingen (Matthias Hein)
Focus areas
- ML for science
- Trustworthy / safe AI
- Robotics learning
- Computer vision
- Efficient ML
How to engage
- Grants / calls (CZS Durchbrüche, CZS Focus, endowed professorships)
- Partner on CZS-funded university projects
- CZS endowed professorships (HAW)
Sources · verified November 2026
Anchor: #eu-carl-zeiss-foundation
CISPA Helmholtz Center for Information Security
Companies needing ML-security, privacy-preserving ML, federated optimisation or LLM-robustness partners and hires; expect extra due diligence while leadership is in transition.
Medium confidenceDetailsClose
National Helmholtz research centre for cybersecurity, privacy and trustworthy AI, founded 2011 and a Helmholtz member since 2019. Its ML faculty produce Europe-leading output on secure, private and efficient learning.
Strengths
- #9 European ML-core output in CSRankings 2021–2025 despite being a security-focused centre
- Rare concentration of ML-security and privacy expertise; 705 A*-venue papers reported by CISPA
- Institutional start-up support via the CISPA Founders Fellowship
- Saarbrücken cluster (Saarland University, MPI-INF/SWS, DFKI) within the ELLIS Unit Saarbrücken
Watch-outs
- Governance turbulence: founding director Michael Backes and CISPA 'mutually agreed to part ways' (Sept 2026) after a special audit; the management page names only COO Kevin Streit
- 2026 audit found insufficient protection of data/trade secrets in China-related cooperation; partners may face added due diligence
- Mission is security-first; ML breadth outside trustworthy/efficient ML is limited
Leading researchers
- FacultySparse/efficient deep learning; top CISPA ML-core output in CSRankings 2021-25
- Sebastian U. StichTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (CISPA; AI & Image Processing; CISPA #1). Distributed and federated optimisation
- Mario FritzTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (CISPA; AI & Image Processing). Trustworthy ML, privacy and generative-model security
- FacultyPrivacy-preserving ML and model-stealing defences
- Website
- cispa.de
- Kind
- Research institute
- Institution
- Helmholtz Association
- Location
- Saarbrücken, Germany
- Founded
- 2011
- CSRankings signal
- #9 in Europe for ML core (ICML/NeurIPS/ICLR), 46.0 adj. pubs, 15 faculty, 2021–2025 (CSRankings data)
- Funding scale
- Helmholtz programme-funded national centre (federal + Saarland); specific budget not stated in sources reviewed
Focus areas
- Trustworthy / safe AI
- Efficient ML
- Theory
- Causality
How to engage
- PhD (with Saarland University)
- Industry partnership programme
- Spin-out / tech transfer (CISPA Founders Fellowship)
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-cispa-helmholtz
Computer Vision Center (CVC), Universitat Autònoma de Barcelona
Automotive and mobility firms needing simulation or perception partners, document-AI and vision-language projects, and companies seeking applied computer-vision transfer or PhD hires in Catalonia.
Medium confidenceDetailsClose
University-based computer-vision institute set up in 1994 by the Catalan government and UAB, with groups in continual learning, vision-and-language/document AI, autonomous driving and human behaviour analysis; it co-hosts the open-source CARLA driving simulator and is one of ten institutions in ELLIS Unit Barcelona (co-directed by CVC's Karatzas).
Strengths
- Co-host of CARLA, a widely adopted open-source driving simulator backed by NVIDIA and Toyota Research Institute
- Three CVC-linked researchers in the 2026 Stanford–Elsevier Spain ICT top-100 view (van de Weijer, Karatzas, Escalera)
- Explicit tech-transfer mission and three-decade track record (est. 1994) with Catalan government backing
- Co-directs ELLIS Unit Barcelona (Karatzas, with Carme Torras of IRI) within a 10-institution Barcelona unit
Watch-outs
- Vision-centred: limited core-ML theory, RL or LLM-pretraining capacity
- Not covered by CSRankings; staff size, budget and director are not stated on its homepage
- Joint appointments across UAB, UB and CVC can complicate IP and contracting
Leading researchers
- Joost van de WeijerTop 2%Senior researcher, CVC; leads the Learning and Machine Perception (LAMP) team; ELLIS Fellowlisted on the topcitedscientists.com Spain ICT page (Centre de Visió per Computador). Continual learning and colour/vision
- Dimosthenis KaratzasTop 2%Professor, UAB; Associate Director, CVC; leads the Vision and Language group; ELLIS Unit Barcelona co-director (with Carme Torras, IRI CSIC-UPC)listed on the topcitedscientists.com Spain ICT page (UAB). Document AI and scene-text understanding
- Lecturer, Dept. of Computer Science, UAB; leads CVC's AI-for-autonomous-driving line; ICREA Academia (2018, renewed 2023); CARLA co-founderTeam created SYNTHIA and UrbanSyn datasets and co-developed CARLA with Intel Labs
- Sergio EscaleraTop 2%Full Professor, Universitat de Barcelona; leads HuPBA (UB–CVC); two-time ICREA Academia awardee; Distinguished Professor, Aalborg Universitylisted on the topcitedscientists.com Spain ICT page (Universitat de Barcelona). Human pose and behaviour analysis
- Website
- cvc.uab.es
- Kind
- Research institute
- Institution
- Generalitat de Catalunya and Universitat Autònoma de Barcelona (UAB)
- Location
- Barcelona (Bellaterra, UAB campus), Spain
- Founded
- 1994
- CSRankings signal
- Not ranked: neither CVC nor UAB is listed as an institution in CSRankings, so output is uncounted (CSRankings data, 2026-11)
Flagship outputs
Focus areas
- Computer vision
- Robotics learning
- NLP / LLMs
- Efficient ML
How to engage
- Industry contract research / tech transfer
- PhD (with UAB and UB) / ELLIS PhD programme
- Open-source collaboration (CARLA)
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-cvc-barcelona
Czech Technical University in Prague AI (CIIRC & Center for Machine Perception; ELLIS Unit Czechia)
Partners in geometric vision, visual localisation and 3D mapping, image matching and tracking, aerial multi-robot systems and Industry 4.0 testbeds; PhD and engineer hiring in Central Europe.
Medium confidenceDetailsClose
CTU's AI strength sits in the Center for Machine Perception (Faculty of Electrical Engineering) and CIIRC, an institute founded in 2013 with eight departments from AI to industrial production. CTU anchors the ELLIS Unit Czechia, added to the network in May 2026.
Strengths
- Top-10 European vision output (#10, CSRankings 2021–2025) from a long-established geometric-vision school
- Strong field robotics: #7 in Europe for robotics venues and an open-source aerial multi-robot stack
- Industry-facing infrastructure at CIIRC: Testbed for Industry 4.0 and national centres for industry, construction and mobility
- Anchors the new ELLIS Unit Czechia with 6 CTU-affiliated Fellows and Scholars listed
Watch-outs
- Weak ML-core output (#73 in Europe) relative to vision and robotics
- CIIRC's breadth (energy, smart cities, biomedical) dilutes its AI identity; staff and budget not stated on its site
- Talent mobility: word2vec author Tomáš Mikolov, founding head of CIIRC's Foundational AI group, is now listed as an alumnus
Leading researchers
- Jiri MatasTop 2%Professor, Center for Machine Perception, CTU; Editor-in-Chief of IJCV; ELLIS Fellowlisted on the topcitedscientists.com Czech Technical University page (AI & Image Processing). RANSAC variants, tracking; 2026 IAPR King-Sun Fu Prize
- Josef SivicTop 2%Researcher, CIIRC CTU (ELLIS also lists ENS and Inria); director of ELLIS Unit Czechialisted on the topcitedscientists.com Czech Technical University page (AI & Image Processing). Visual recognition and video understanding
- Torsten SattlerTop 2%Senior Researcher, head of Spatial Intelligence group, CIIRC (since July 2020); ELLIS Fellowlisted on the topcitedscientists.com Czech Technical University page (AI & Image Processing). Visual localisation, 3D mapping, neural rendering
- Website
- ciirc.cvut.cz
- Kind
- University group / lab
- Institution
- Czech Technical University in Prague (CTU)
- Location
- Prague, Czechia
- Founded
- 2013
- CSRankings signal
- #10 in Europe for vision (22.6 adj.), #7 for robotics (ICRA/IROS/RSS), #73 for ML core (4.2 adj.), 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Computer vision
- Robotics learning
- Theory
- Trustworthy / safe AI
How to engage
- PhD / ELLIS PhD programme
- Industry testbeds (Testbed for Industry 4.0, RICAIP, national 4.0 centres)
- Research incubator (eClub) / spin-outs
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-ctu-prague-ciirc-cmp
DFKI – German Research Center for Artificial Intelligence
Industrial firms and SMEs seeking applied, contract-ready AI R&D (robotics, language technology, manufacturing, health) or a partner for publicly co-funded German/EU projects; not the first stop for frontier ML research.
Medium confidenceDetailsClose
Non-profit public-private partnership founded in 1988 and one of the largest applied-AI research centres in Germany, with sites across eight cities and funding from the EU, federal ministries, states, DFG and industry.
Strengths
- Applied, contract-ready research across robotics, language technology, vision, HCI and IT security (Wikipedia)
- Public-private structure (industrial and public shareholders) designed for industry-ready contract research
- Nationwide presence: HQ Kaiserslautern plus Saarbrücken, Bremen, Oldenburg, Osnabrück and labs in Berlin, Darmstadt, Lübeck
- Diverse public funding base (EU, BMFTR, BMWK, states, DFG) alongside industrial partners
Watch-outs
- Weak academic-impact signal: not listed in CSRankings, and only 7 DFKI-affiliated authors appear in the 2026 Stanford–Elsevier list (neither of the two researchers named here)
- Large, multi-site organisation; quality and focus vary by department, so partners must vet the specific team
- Not an ELLIS Unit participant in sources reviewed; headcount and budget not stated on the pages reviewed
Leading researchers
- CEO (with CFO Helmut Ditzer)Human-computer interaction and ubiquitous AI; leads DFKI
- Researcher (DFKI profile; CSRankings-listed at Saarland University)Machine translation and multilingual NLP
- Website
- dfki.de
- Kind
- Research institute
- Location
- Kaiserslautern (HQ); Saarbrücken, Bremen, Oldenburg, Osnabrück; labs in Berlin, Darmstadt, Lübeck; branch in Trier, Germany
- Founded
- 1988
- CSRankings signal
- Not listed as an institution in CSRankings (DFKI researchers counted under partner universities)
- Funding scale
- Mixed funding: EU, BMFTR, BMWK, federal states, DFG and industrial partners (total not stated)
Focus areas
- Robotics learning
- NLP / LLMs
- Computer vision
- Trustworthy / safe AI
How to engage
- Contract research
- Publicly funded consortium projects (EU, BMFTR, BMWK)
- Spin-out / tech transfer
- Living labs / testbeds
Sources · verified November 2026
Anchor: #eu-dfki
Dieter Schwarz Foundation (Dieter Schwarz Stiftung) & IPAI Heilbronn
International AI postdocs (DAAD–DSS Heilbronn programme), researchers considering endowed chairs at TUM/ETH Heilbronn, and companies joining the IPAI ecosystem.
Medium confidenceDetailsClose
Private foundation of Lidl/Schwarz Group founder Dieter Schwarz (active since 1999); funds 41 TUM professorships (32 in Heilbronn, incl. 10 data-science/AI chairs added in 2022), 20 ETH Zurich professorships over 30 years, and backs the IPAI AI innovation park in Heilbronn.
Strengths
- Very long-horizon commitments: TUM and ETH chairs financed for ~30 years incl. equipment
- Builds a full ecosystem in Heilbronn: TUM and ETH campuses, Fraunhofer centres, IPAI park
- New 2026 DAAD partnership offers concrete postdoc/early-career routes for international AI talent
Watch-outs
- Does not publish its budget or the financial volume of its partnerships
- Geographically concentrated on Heilbronn-Franken; no open AI grant calls
- Close ties to Schwarz Group business interests (e.g., IPAI's stake in Aleph Alpha's 2023 round) raise independence questions
Leading researchers
- Managing Director, Dieter Schwarz FoundationQuoted on the 2026 DAAD–DSS international AI talent programme
- Managing director for science (as reported Jan 2024)Led the foundation's science and AI partnerships (TUM, ETH, Aleph Alpha)
- Website
- dieter-schwarz-stiftung.de
- Kind
- Foundation / funder
- Location
- Heilbronn / Neckarsulm, Germany
- Founded
- 1999
- Funding scale
- Budget undisclosed; ETH Zurich gift of 20 professorships for 30 years estimated at several hundred million CHF (Dec 2023); 41 TUM chairs funded for an initial 30 years; foundation says its research funding exceeds VolkswagenStiftung's (~€100m/yr)
Flagship outputs
- DAAD–DSS 'Attracting International AI Talent to Heilbronn' (announced 23 Apr 2026; 3-year pilot, up to 70 talents/year; first call opened 1 May 2026)
- 41 DSS-funded TUM professorships (32 in Heilbronn), incl. 10 data-science/AI chairs added Oct 2022, financed for an initial 30 years
- IPAI campus, Heilbronn: planned 30-ha AI district for 5,000+ people (groundbreaking Nov 2025, first buildings due 2027); IPAI platform has 80+ companies and institutions
Focus areas
- NLP / LLMs
- Trustworthy / safe AI
- Efficient ML
How to engage
- Fellowships (DAAD–DSS international AI talent programme)
- Endowed professorships (via TUM / ETH Zurich)
- IPAI membership (companies)
- Strategic philanthropy
Sources · verified November 2026
Anchor: #eu-dieter-schwarz-foundation-ipai
ELLIS Alicante – Institute of Humanity-centric AI
Partners interested in responsible/human-centric AI, AI and wellbeing, and algorithmic bias; PhD and postdoc candidates via the ELLIS PhD programme; sponsors seeking a Spanish ELLIS unit.
Medium confidenceDetailsClose
Non-profit private research foundation (constituted May 2020) and the first ELLIS unit in Spain, created from scratch; researches human-centric, trustworthy AI (behaviour modelling, human–AI interaction, bias, AI for social good).
Strengths
- Clear niche: one of few European ML institutes dedicated to human-centric and socially beneficial AI
- Director is a highly decorated researcher (ACM/IEEE Fellow, Academia Europaea, 2024 Hypatia European Science Prize)
- International scientific advisory board (set up Dec 2021) incl. Krishna Gummadi, Adrian Weller and Sandy Pentland
- Only ELLIS unit built from scratch in Spain; can host ERC grantees
Watch-outs
- Small institute; public pages give no headcount or annual budget
- Relies on regional government (and ERDF) support plus sponsorship; in May 2021 no private sponsorship deals had yet been closed
- Limited core-ML (ICML/NeurIPS/ICLR) output relative to large ELLIS units
Leading researchers
- Co-founder and Director, ELLIS AlicanteHuman behaviour modelling and human–machine interaction; ACM and IEEE Fellow; 2025 Julio Rey Pastor National Research Award
- Website
- ellisalicante.org
- Kind
- Research institute
- Institution
- Fundación de la Comunitat Valenciana Unidad ELLIS Alicante (ELLIS network)
- Location
- Alicante, Spain
- Founded
- 2020
- Funding scale
- Generalitat Valenciana: €1.5m in its 2021 budget (ELLIS hosts must commit ≥5 years and ≥€1.5m/yr); partly funded by the Generalitat Valenciana and the EU ERDF (per homepage); private sponsors not itemised on fetched pages
Flagship outputs
Focus areas
- Trustworthy / safe AI
- NLP / LLMs
- Computer vision
How to engage
- PhD / ELLIS PhD programme
- Host for ERC grants and ValER positions
- Sponsorship & donations
- Strategic partnership
Sources · verified November 2026
Anchor: #eu-ellis-alicante
ELLIS Institute Tübingen
AI-safety, AutoML and foundation-model partners seeking independent PIs on long-horizon endowed positions; funders backing open European AI; recruiters targeting the ELLIS PhD programme in Tübingen.
High confidenceDetailsClose
Europe's first ELLIS Institute, launched July 2023 with €100m from the Hector Foundation plus Baden-Württemberg support. It hosts 14 research groups, most of whose PIs are co-affiliated with MPI-IS.
Strengths
- Large private endowment (€100m) allows long-horizon, industry-competitive PI positions
- Concentration of AI-safety and alignment PIs; three PIs named 2026 Schmidt Sciences AI2050 Fellows
- Tight integration with MPI-IS and the Tübingen Cluster of Excellence
- Spin-out track record via PIs: Prior Labs (TabPFN), which SAP agreed to acquire in 2026 with a >€1bn investment plan
Watch-outs
- Young institution (2023) with a small number of groups; outputs still maturing
- Many PIs hold dual affiliations, so contracting counterparty may be MPI-IS or a university
- Dependence on a single private donor for core funding
Leading researchers
- Bernhard SchölkopfTop 2%PI, Empirical Inference; Scientific Director at the 2023 launchin the V9 (Aug 2026) list per the topcitedscientists.com mirror (MPI for Intelligent Systems; AI & Image Processing; global rank #475). Causal ML; ELLIS president at launch
- Frank HutterTop 2%Hector Endowed Fellow and PI (AutoML); co-founder & CEO of Prior Labsin the V9 (Aug 2026) list per the topcitedscientists.com mirror (affiliation listed as Prior Labs; Germany ICT #14). AutoML; led TabPFN (Nature 2025); co-founder and CEO of Prior Labs, which SAP agreed to acquire in May 2026 (tech.eu reported the deal done in July 2026)
- PISafety- and efficiency-aligned learning
- Maksym AndriushchenkoTop 2%PIin the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under MPI for Intelligent Systems; AI & Image Processing). AI safety and alignment; LLM robustness
- PIAlgorithms and society; performative prediction
- Website
- institute-tue.ellis.eu
- Kind
- Non-profit open lab
- Location
- Tübingen, Germany
- Founded
- 2023
- CSRankings signal
- Not listed separately in CSRankings; PIs are counted under Max Planck Society, Tübingen or Freiburg
- Funding scale
- €100m Hector Foundation endowment (2023) plus Baden-Württemberg state funding
Flagship outputs
Focus areas
- Trustworthy / safe AI
- Efficient ML
- NLP / LLMs
- ML for science
- Causality
How to engage
- PhD / ELLIS PhD programme
- Hector Endowed Fellow PI positions
- Industry partnership programme
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-ellis-institute-tuebingen
ELLIS – European Laboratory for Learning and Intelligent Systems
Anyone mapping Europe's top ML talent: use ELLIS Units to find PIs, the ELLIS PhD & Postdoc programme (annual call; the 2027-intake deadline was 31 Oct 2026) to recruit, and sponsorship to reach the network.
High confidenceDetailsClose
Pan-European ML network of excellence launched at NeurIPS in December 2018, now spanning 52 sites (49 Units plus Institutes in Tübingen and Finland) in 21 countries, 701 Fellows & Scholars and 2.3K+ members.
Strengths
- Best single map of European ML excellence: 701 Fellows & Scholars across 52 sites
- Central PhD/postdoc application covering 50–100 institutions with cross-border co-supervision
- 16 cross-border research programmes led by Fellows
- Spawned endowed ELLIS Institutes (Tübingen 2023, Finland 2025)
Watch-outs
- A network, not an employer: contracts and IP sit with member institutions
- Unit pages vary in detail; membership data is uneven across sites
- Includes non-European units (e.g., Tel Aviv), so not a pure Europe filter
Leading researchers
- Bernhard SchölkopfTop 2%Co-founder & former Presidentin the V9 (Aug 2026) list per the topcitedscientists.com mirror (MPI for Intelligent Systems; AI & Image Processing; global rank #475). Causal ML; co-founder of ELLIS; listed on ellis.eu as "Co-Founder & Former President"
- Sepp HochreiterTop 2%Board guestin the V9 (Aug 2026) list per the topcitedscientists.com mirror (JKU Linz; AI & Image Processing; global rank #117). LSTM co-inventor; JKU Linz
- Board memberLed ELLIS Unit Alicante (foundation-run unit) at its 2020 launch
- Website
- ellis.eu
- Kind
- Network / society
- Location
- Pan-European, Europe (21 countries), Pan-European
- Founded
- 2018
Flagship outputs
Focus areas
- Theory
- Probabilistic ML
- Computer vision
- NLP / LLMs
- Robotics learning
- ML for science
How to engage
- PhD / ELLIS PhD programme
- Sponsorship & donations
- ELLIS Units (local collaboration)
- Summer/winter schools
Sources · verified November 2026
Anchor: #eu-ellis-society
EMBL AI – European Molecular Biology Laboratory (AI Hub Heidelberg & EMBL-EBI)
ML researchers moving into biology (group-leader and postdoc posts), teams needing open biodata for model training, and companies seeking EMBL-EBI data and industry collaborations in AI for life sciences.
Medium confidenceDetailsClose
Europe's intergovernmental life-sciences lab, now building 'EMBL AI' under a Feb 2025 AI science strategy: a new AI Hub in Heidelberg with AI groups and engineering teams, plus EMBL-EBI's ML-ready biodata (AlphaFold DB).
Strengths
- Unique data position: EMBL-EBI runs core open biodata resources incl. AlphaFold DB (200m+ structures)
- Dedicated philanthropic backing: €40m+ from the Hector Foundation for AI groups, engineers and fellowships
- Part of ELLIS Unit Heidelberg: Stegle is a unit director, with Huber (Fellow) and Kreshuk (Scholar) as members
- Recurring, international fellowship routes (EIPOD-LinC) with an explicit AI and theory theme
Watch-outs
- AI Hub is new (2025–26): leadership is still 'acting' and many posts were only recently advertised
- Compute investment is unspecified in the strategy; relies partly on national/European supercomputers
- Biology-first agenda: limited fit for general ML research without a life-science question
Leading researchers
- Acting Head of EMBL AI (per 2025 article and a 2026 job posting); director of ELLIS Unit Heidelberg (Fellow; EMBL and DKFZ)Statistical ML for genomics and single-cell biology; co-chaired EMBL's AI science strategy
- Senior scientist, EMBL Heidelberg; ELLIS ScholarML for bioimage analysis (e.g., PlantSeg deep-learning segmentation)
- Group leader, EMBL Heidelberg; ELLIS FellowStatistical computing for genomics; member of the EMBL AI science working group
- Team leader, EMBL-EBI (PDBe and AlphaFold DB)Oversees the Protein Data Bank in Europe and the AlphaFold Protein Structure Database
- Website
- embl.org
- Kind
- Research institute
- Institution
- European Molecular Biology Laboratory (intergovernmental organisation, 29 member states)
- Location
- Heidelberg (AI Hub, HQ); Hinxton (EMBL-EBI); also Barcelona, Grenoble, Hamburg, Rome, Germany
- Founded
- 1974
- Funding scale
- Intergovernmental core funding from 29 member states (AI share not broken out); over €40m long-term gift from the Hector Foundation for EMBL AI (Jul 2025)
Flagship outputs
- AlphaFold Protein Structure Database (with Google DeepMind): 200m+ predicted structures (header now says 260m+), CC-BY-4.0
- EMBL AI Science Strategy (Feb 2025): methodology, data-to-insight and lab-in-the-loop pillars
- EIPOD-LinC postdoctoral fellowships incl. an 'AI and Theory' area (MSCA Cofund; 2026 call open 25 Nov 2026–1 Feb 2027)
Focus areas
- ML for science
- Computer vision
- Probabilistic ML
- Causality
How to engage
- Postdoctoral fellowships (EIPOD-LinC, 3 years, partner lab in an EMBL member state)
- Group-leader, team-leader and AI-engineer hiring for the AI Hub
- EMBL International PhD Programme
- Industry collaboration (e.g., Google DeepMind on AlphaFold DB; Open Targets model)
- Courses and conferences (EMBL-EBI AI training; EMBO|EMBL AI and biology symposium)
Sources · verified November 2026
Anchor: #eu-embl-ai
ENS-PSL & Collège de France: mathematics of machine learning (ELLIS Unit Paris host)
Partners seeking mathematical foundations of ML (optimal transport, wavelets/scattering, optimisation, statistical physics of learning), PhD/postdoc hiring in Paris, and start-ups drawing on ENS's lab spin-out record. For chair/fellow funding calls see the prairie-psai entry.
Medium confidenceDetailsClose
Mathematics-of-learning research at École normale supérieure (computer science and mathematics departments, ENS Data Science Center created 2020) and Collège de France, within Université PSL: optimal transport, wavelets/scattering and optimisation. ENS directs ELLIS Unit Paris (Peyré). Many of these researchers hold chairs in the PSL-led PR[AI]RIE-PSAI cluster, which is listed separately (see prairie-psai) and is where its €75m award and recruitment calls are described.
Strengths
- Strong spin-out record: ENS reports 15+ AI start-ups from its labs since 2015, incl. Mistral AI, Dataiku and Shift Technology
- World-class mathematics-of-learning school: optimal transport (Peyré), scattering (Mallat), optimisation (d'Aspremont)
- Several researchers in the 2026 Stanford–Elsevier lists (AI & Image Processing): Bach and Peyré (PSL page), Mallat (Collège de France), d'Aspremont (ENS)
- ELLIS Unit Paris is directed from ENS (Peyré), linking PSL to Hi! PARIS, SCAI and DATAIA
Watch-outs
- Heavy overlap with Inria and CNRS staff (several chairs hold Inria posts), which blurs who contracts and owns IP
- Bibliometric visibility is low in CSRankings (ENS #47 for ML core) because staff publish under many affiliations
- No single ENS AI institute or contact point: work is spread across DI, DMA, the Data Science Center, Collège de France and Inria teams
Leading researchers
- Gabriel PeyréTop 2%CNRS senior researcher, ENS (DMA); director ENS Center for Data Science and ELLIS Unit Paris; PR[AI]RIE-PSAI management committeelisted on the topcitedscientists.com Université PSL page (AI & Image Processing). Optimal transport; ERC Advanced Grant WOLF
- Stéphane MallatTop 2%Professor, Collège de France (Chair of Data Sciences); ELLIS Fellow (listed at ENS)listed on the topcitedscientists.com Collège de France page (AI & Image Processing). Wavelets and scattering networks
- Alexandre d'AspremontTop 2%CNRS researcher, ENS Computer Science Dept. (Inria SIERRA team); co-founder of Kayrroslisted on the topcitedscientists.com École Normale Supérieure page (AI & Image Processing). Convex optimisation
- Website
- ai-inthecity.ens.psl.eu
- Kind
- University group / lab
- Institution
- Université PSL (with Université Paris Cité, CNRS, Inria, Institut Pasteur)
- Location
- Paris, France
- CSRankings signal
- ENS: #47 in Europe for ML core (7.1 adj., 5 listed faculty), 2021–2025; CSRankings lists few CNRS/Inria staff, so PSL output is undercounted (CSRankings data)
- Funding scale
- No dedicated budget published for these groups; many researchers are funded through PR[AI]RIE-PSAI chairs (see prairie-psai for the €75m IA Cluster award), CNRS/Inria posts and ERC grants (e.g. Peyré's ERC Advanced Grant WOLF)
Flagship outputs
Focus areas
- Theory
- ML for science
- Computer vision
- Probabilistic ML
How to engage
- PhD / ELLIS PhD programme
- Industry partnership via the PR[AI]RIE-PSAI cluster (see prairie-psai)
- Spin-out / tech transfer (ENS reports 15+ AI start-ups from its labs since 2015)
- AI Fellow and chair positions (PSL recruitment calls; details in prairie-psai)
Sources · verified November 2026
Anchor: #eu-psl-ens-prairie-psai
EPFL AI Center
Companies seeking research partners or hires in optimisation, learning theory, efficient/distributed training and ML for health, and teams wanting to build on Swiss AI Initiative open models (Apertus); engage via the Industry Partnership Program or individual labs.
High confidenceDetailsClose
EPFL's cross-school AI hub, launched in 2023, with 100+ affiliated professors and Amina Chebira as Executive Director. With ETH Zurich and CSCS it co-runs the Swiss AI Initiative behind the open Apertus LLM.
Strengths
- Highest ML-core publication volume of any European institution in CSRankings (2021–2025), across 38 listed faculty
- Co-developer of Apertus, a fully open 8B/70B multilingual LLM released September 2025
- Access to national-scale compute via CSCS Alps (10,000+ GH200 GPUs) through the Swiss AI Initiative
- Strong theory and optimisation bench (Cevher, Flammarion, Krzakala, Zdeborová) alongside applied health AI (MEDITRON)
Watch-outs
- The Center coordinates 100+ professors rather than employing them; contracts and IP are negotiated lab-by-lab with EPFL's tech-transfer office
- Relatively weak in classic AI venues (AAAI/IJCAI rank #17 in Europe) versus its ML lead
- Swiss non-EU status can complicate some EU-funded consortium roles and data-transfer terms
Leading researchers
- Pascal FrossardTop 2%Co-director at the Center's 2023 launchin the V9 (Aug 2026) list per the topcitedscientists.com mirror (EPFL; Swiss ICT list #73). Signal processing and ML for graphs; co-leads the Center
- Marcel SalathéTop 2%Co-director at the Center's 2023 launchin the V9 (Aug 2026) list per the topcitedscientists.com mirror (EPFL; AI & Image Processing). Digital epidemiology and AI for health; co-leads the Center
- Volkan CevherTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (EPFL; Swiss ICT #84; topscinet shows 2025 and 2026 single-year, AI&IP subfield rank 1,727 (2026)). Optimisation for ML; 4th-highest ML-core output of any European academic in CSRankings 2021-25
- Martin JaggiTop 2%Professor of Machine Learningin the V9 (Aug 2026) list per the topcitedscientists.com mirror (EPFL MLO lab; Swiss ICT list #2). Distributed/efficient ML; on Swiss AI Initiative steering committee (Apertus)
- Lenka ZdeborováTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (EPFL; subfield Fluids & Plasmas). Statistical physics of learning; high ML-core output in CSRankings data
- Website
- ai.epfl.ch
- Kind
- Research institute
- Institution
- EPFL (École polytechnique fédérale de Lausanne)
- Location
- Lausanne, Switzerland
- Founded
- 2023
- CSRankings signal
- #1 in Europe for ML core (ICML/NeurIPS/ICLR), 2021–2025: 109.6 adjusted pubs, 38 faculty; #1 on ML+AI+CV+NLP combined (CSRankings data)
- Funding scale
- Swiss AI Initiative (shared with ETH): CHF 20m ETH Domain grant plus 10m+ GPU hours on Alps
Flagship outputs
Focus areas
- Theory
- Efficient ML
- NLP / LLMs
- Computer vision
- ML for science
- Trustworthy / safe AI
How to engage
- PhD / ELLIS PhD programme
- Postdoctoral fellowships (EPFL AI Center calls)
- Industry partnership programme
- Spin-out / tech transfer
- Grants / calls (Swiss AI Initiative large-project calls)
Sources · verified November 2026
Anchor: #eu-epfl-ai-center
ETH AI Center
Companies and investors wanting co-supervised doctoral/postdoctoral AI fellows, collaboration on trustworthy AI, LLM security and probabilistic ML, and access to the Zurich spin-out pipeline; also Swiss AI Initiative partners.
High confidenceDetailsClose
ETH Zurich's interdisciplinary AI hub, established in 2020 and spanning 110+ professorships across all 16 departments. It runs doctoral and postdoctoral fellowship programmes and co-leads the Swiss AI Initiative.
Strengths
- Breadth: 110+ affiliated professorships across all 16 ETH departments plus UZH and HSG groups
- Top-3 European ML output and #2 in NLP in CSRankings 2021–2025
- Fellowship model pairs each fellow with two mentors from different fields
- Co-leads Swiss AI Initiative and Apertus with national supercomputing access (Alps)
Watch-outs
- Highly competitive fellowship intake (5–8 doctoral places per year)
- Center coordinates rather than employs most PIs, so IP/contracting goes through ETH units
- Weaker relative standing in classic AI venues (AAAI/IJCAI rank #50 in Europe)
Leading researchers
- Andreas KrauseTop 2%Professor; chair of the AI Center steering committeein the V9 (Aug 2026) list per the topcitedscientists.com mirror (ETH Zürich; AI & Image Processing; Swiss ICT #10). Probabilistic ML and RL; 3rd-highest European ML-core output in CSRankings 2021-25
- Co-founder & Executive DirectorRuns the Center's fellowship, start-up and partnership activities
- Ryan CotterellTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (ETH Zürich; Swiss ICT #95). NLP / computational linguistics; drives ETH's #2 European NLP ranking
- Florian TramèrTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (ETH Zürich; AI & Image Processing; Swiss ICT #7). Security and privacy of ML systems, incl. LLM attacks
- Martin VechevTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (ETH Zürich; Swiss ICT #65). Reliable/secure AI and program analysis; also founded INSAIT (Sofia)
- Website
- ai.ethz.ch
- Kind
- Research institute
- Institution
- ETH Zurich
- Location
- Zurich, Switzerland
- Founded
- 2020
- CSRankings signal
- #3 in Europe for ML core (ICML/NeurIPS/ICLR), 2021–2025: 91.4 adjusted pubs; #2 for NLP; #3 on ML+AI+CV+NLP (CSRankings data)
- Funding scale
- Doctoral fellows paid CHF 72.8k–83.2k/yr; Swiss AI Initiative CHF 20m ETH Domain grant (shared with EPFL)
Flagship outputs
Focus areas
- Probabilistic ML
- RL
- NLP / LLMs
- Trustworthy / safe AI
- Computer vision
- Robotics learning
How to engage
- PhD / ELLIS PhD programme
- AI Center doctoral & postdoctoral fellowships
- Industry partnership programme
- Spin-out / tech transfer
- Grants / calls (Swiss AI Initiative)
Sources · verified November 2026
Anchor: #eu-eth-ai-center
EuroHPC AI Factories (EuroHPC Joint Undertaking)
European academic labs needing free GPU time for open research, AI startups/SMEs needing entry-level to 50k+ GPU-hour allocations, and EU-funded project consortia.
High confidenceDetailsClose
EU network of 19 AI Factories plus 13 AI Factory Antennas built around AI-optimised EuroHPC supercomputers, offering peer-reviewed GPU allocations; free for publicly funded research and for AI SMEs/startups on innovation work.
Strengths
- Largest public AI compute offer in Europe: 19 AI Factories in 16+ countries plus 13 Antennas (incl. UK and Switzerland)
- Free GPU time for publicly funded research; free Playground access for AI SMEs and startups
- Fast, continuous access: Large Scale has two cut-offs per month with decisions in ~10 working days
Watch-outs
- Demand exceeds supply: MareNostrum and LUMI-G unavailable for new Large Scale allocations, and MareNostrum 5 ACC unavailable for Playground access (as of Nov 2026)
- Critics (CLAIRE co-founder Holger Hoos, Jan 2024) argued the factory model spreads money and compute too thinly versus a 'CERN for AI'
- Compute only: peer review, eligibility rules and support quality vary across host sites
Leading researchers
- Executive Director, EuroHPC JU (since Sep 2020)Oversees the JU's supercomputers and the AI Factories programme
- Website
- eurohpc-ju.europa.eu
- Kind
- Network / society
- Institution
- EuroHPC Joint Undertaking (EU, Member States and Associated Countries)
- Location
- Luxembourg (JU seat); AI Factories across Europe, Luxembourg (EU-wide network), Pan-European
- Founded
- 2024
- Funding scale
- ~€10bn EU + national investment in EuroHPC supercomputing and AI Factories (2021–2027); AI Gigafactories: up to €10bn EU and national funding (Commission), with >€20bn private investment expected (JU)
Flagship outputs
- AI Factories access modes: Playground, Fast Lane (≤50k GPU-h), Large Scale (>50k GPU-h), AI for Science (free)
- Large Scale access: continuously open, two cut-offs per month, decisions within 10 working days
- AI Gigafactories call (launched 30 Jul 2026, deadline 12 Nov 2026): up to seven facilities in at least seven Member States
Focus areas
- NLP / LLMs
- Efficient ML
- ML for science
How to engage
- Compute access (AI for Science and Collaborative EU projects, free)
- Compute access for industry (Playground, Fast Lane, Large Scale)
- AI Factory support services (training, data, expertise)
- AI Gigafactory consortia (tender closes 12 Nov 2026)
Sources · verified November 2026
Anchor: #eu-eurohpc-ai-factories
European Research Council (ERC)
Researchers relocating to or based in Europe seeking multi-year investigator grants; founders and investors use ERC grants as a signal of PI quality and PoC grants as a spin-out precursor.
High confidenceDetailsClose
The EU's frontier-research funder (>€16bn for 2021–2027, 17% of Horizon Europe), awarding Starting, Consolidator, Advanced, Synergy, Proof of Concept and (from 2026) ERC Plus grants to researchers of any nationality.
Strengths
- Largest frontier-research funder in Europe; open to any nationality
- ERC grants are a strong quality signal when vetting academic PIs and spin-out teams
- Proof of Concept scheme supports commercialisation of ERC results
- Many leading European ML PIs hold ERC grants (e.g., Hutter, De Raedt, Van Gool)
Watch-outs
- Highly competitive (e.g., 3,458 Advanced Grant proposals in 2026)
- Funds individual investigators, not companies or consortia
- Not AI-specific; ML competes with all fields in its panels
Leading researchers
- ERC President (since Nov 2021; re-appointed Oct 2025 to end-Sept 2027)Leads the ERC Scientific Council
- Website
- erc.europa.eu
- Kind
- Foundation / funder
- Institution
- European Commission (Horizon Europe)
- Location
- Brussels, Belgium (EU-wide), Pan-European
- Founded
- 2007
- Funding scale
- >€16bn for 2021–2027; 2026 Starting Grants €705m (421 grants)
Flagship outputs
Focus areas
- Theory
- ML for science
- Trustworthy / safe AI
- Computer vision
- NLP / LLMs
How to engage
- Grants / calls (StG, CoG, AdG, SyG, PoC, ERC Plus)
- Host an ERC grantee (institutions)
Sources · verified November 2026
Anchor: #eu-european-research-council
Finnish Center for AI (FCAI) & ELLIS Institute Finland
Partners seeking probabilistic and data-efficient ML, human-AI collaboration research or PI-level hires in Finland, and access to LUMI-scale compute; engage via FCAI/ELLIS Institute Finland calls.
High confidenceDetailsClose
FCAI is a Research Council of Finland flagship initiated by Aalto, Helsinki and VTT; it forms the foundation of ELLIS Institute Finland, established 2025 with €10m/yr state funding (2025–2028).
Strengths
- New national-scale ELLIS Institute with committed multi-year state and private funding
- Research Council of Finland flagship status for FCAI; all Finnish universities partner as equals
- Long tradition in probabilistic modelling and ICA-style identifiable learning
- Researchers get access to the EuroHPC LUMI supercomputer
Watch-outs
- Institute is newly established (2025); PI recruitment and outputs still ramping up
- Mid-table CSRankings output (Aalto #16 in Europe) compared with Swiss/UK leaders
- State funding committed only to 2028; continuity beyond that not yet public
Leading researchers
- Samuel KaskiTop 2%Director, FCAI and ELLIS Institute Finland (Aalto)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (Aalto; AI & Image Processing). Probabilistic ML and human-in-the-loop AI; led ELLIS Unit Helsinki
- Faculty, AaltoGraph and generative ML; top Aalto ML-core output in CSRankings 2021-25
- Arno SolinTop 2%Faculty, Aaltoin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Aalto; AI & Image Processing). Gaussian processes and probabilistic sensor fusion
- Aapo HyvärinenTop 2%Professor, University of Helsinkiin the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under University of Helsinki; AI & Image Processing; global rank #987). Independent component analysis and identifiable representation learning
- Website
- fcai.fi
- Kind
- Research institute
- Institution
- Aalto University (coordinator), University of Helsinki, VTT
- Location
- Espoo / Helsinki, Finland
- Founded
- 2025
- CSRankings signal
- Aalto: #16 in Europe for ML core (ICML/NeurIPS/ICLR), 25.5 adj., 13 faculty; Univ. of Helsinki #42 (8.2), 2021–2025 (CSRankings data)
- Funding scale
- €10m/yr Ministry of Education and Culture (2025–2028) + €6m start-up + €10m Foundation PS donation
Flagship outputs
Focus areas
- Probabilistic ML
- ML for science
- Trustworthy / safe AI
- Theory
- Efficient ML
How to engage
- PhD / ELLIS PhD programme
- Industry & public-sector partnerships
- PI recruitment (ELLIS Institute Finland calls)
- Spin-out / tech transfer
Sources · verified November 2026
Anchor: #eu-fcai-ellis-institute-finland
Fondazione FAIR – Future Artificial Intelligence Research
Companies and researchers seeking Italian AI partners across many universities, SMEs targeting cascade calls, and recruiters of the 500+ young researchers trained under FAIR.
Medium confidenceDetailsClose
Non-profit foundation (est. Sept 2022, Pisa) coordinating Italy's largest public AI research programme (PNRR extended partnership PE00000013, ~€115m): 10 thematic spokes and 7 transversal projects across 136 universities, research bodies and companies; continues with follow-on PON projects after the PNRR phase ended (Dec 2025).
Strengths
- National scale: 136 participating universities, research bodies and companies across 16 regions
- Trained a large talent pool: 1,033 scientists incl. 516 newly recruited young researchers and PhD students
- Output well above targets: ~3,000 publications by mid-2025 (334% of expected threshold)
Watch-outs
- Core PNRR funding ended in Dec 2025; scale of post-PNRR funding is not public
- Highly distributed structure (hub + 10 spokes) makes it hard to identify a single lab or contact
- Publication counts are self-reported; limited evidence of flagship models or tools comparable to peer programmes
Leading researchers
- Vice President, Fondazione FAIR board (University of Modena and Reggio Emilia)Computer vision researcher; listed on the foundation's governance page
- President, Fondazione FAIR (CNR), per the governance page; reported in Dec 2025 as president of the Technical-Scientific CommitteeDescribed FAIR's main achievement as reducing fragmentation of Italy's AI research network
- Website
- fondazione-fair.it
- Kind
- Network / society
- Institution
- Public-private foundation; hub of Italy's PNRR AI extended partnership (PE00000013, proposed by CNR)
- Location
- Pisa, Italy
- Founded
- 2022
- Funding scale
- €114.5m PNRR (MUR) at award for Jan 2023–Dec 2025 (University of Bologna project page); an updated €116.1m total and a €48.1m Southern-Italy share were reported by Agenda Digitale but could not be re-verified; follow-on PON projects (SINTESI, AI4Nature)
Flagship outputs
Focus areas
- Trustworthy / safe AI
- Efficient ML
- NLP / LLMs
- Computer vision
- Robotics learning
How to engage
- Cascade calls (bandi a cascata) for universities and SMEs
- Hub recruitment
- Joint EU projects via partner network
- Industry partnership (founding member companies)
- Spoke-level collaboration (e.g., Polimi leads Spoke 4 Adaptive AI)
Sources · verified November 2026
Anchor: #eu-fondazione-fair
Fraunhofer IAIS – Institute for Intelligent Analysis and Information Systems
Companies and public bodies wanting contract R&D, AI assessment/assurance, or custom European-language LLMs; partners for German/EU-funded consortium projects.
Medium confidenceDetailsClose
Applied AI institute (~350 staff) of the Fraunhofer-Gesellschaft; led OpenGPT-X (Teuken-7B, 24 EU languages), partners in OpenEuroLLM and the JUPITER AI Factory, and is a lead partner of the Lamarr Institute.
Strengths
- Built one of few European LLMs trained from scratch on all 24 EU languages (Teuken-7B) with custom multilingual tokenizer
- Embedded in major EU/German AI infrastructure: OpenEuroLLM, JUPITER AI Factory, WestAI service centre
- Strong industry interface: contract R&D across finance, health, industry, media and public sector
- Its deputy director manages the Fraunhofer Big Data and AI Alliance, linking IAIS to other Fraunhofer institutes
Watch-outs
- Applied/contract orientation: limited top-venue ML research output compared with university labs
- Teuken-7B is small; at release it led similar-sized open models on average across three benchmarks but placed second on each individual one
- Research-only licence variant alongside the Apache version complicates reuse
Leading researchers
- Director, Fraunhofer IAIS; Professor, University of BonnMachine learning and data mining; quoted as IAIS Director on the Teuken release
- Deputy Director, Fraunhofer IAIS; managing director of the Fraunhofer Big Data and AI Alliance; Professor of Data Analytics, University of SiegenHeads the Hybrid AI research and innovation centre in Heilbronn; Lamarr associated PI
- Website
- iais.fraunhofer.de
- Kind
- Research institute
- Institution
- Fraunhofer-Gesellschaft
- Location
- Sankt Augustin (Bonn), with sites in Dresden and Heilbronn, Germany
- Founded
- 2006
- Funding scale
- Institute budget not stated on fetched pages (~350 staff); led OpenGPT-X, funded by BMWK with ~€14m (Jan 2022–Mar 2025)
Flagship outputs
Focus areas
- NLP / LLMs
- Trustworthy / safe AI
- Efficient ML
How to engage
- Contract research
- AI assessment & assurance services
- Publicly funded consortium projects (EU, federal, NRW)
- Training / certification courses
- Lamarr Institute collaboration
Sources · verified November 2026
Anchor: #eu-fraunhofer-iais
Gatsby Charitable Foundation
Funders and recruiters interested in the theory-heavy ML/neuroscience talent pipeline around UCL's Gatsby Unit; it commissions rather than takes open applications.
Medium confidenceDetailsClose
Foundation set up by David Sainsbury that has committed over £1.7bn since 1967; in ML it funds UCL's Gatsby Computational Neuroscience Unit, combining theoretical neuroscience and machine learning.
Strengths
- Long-horizon funding behind one of the most influential ML/neuroscience units in Europe
- Gatsby Unit alumni include two Google DeepMind co-founders
- Neuroscience portfolio links ML to brain science (UCL and Sainsbury Wellcome Centre)
- Very large cumulative philanthropy (£1.7bn+)
Watch-outs
- Typically commissions work with experts; no open application route for AI stated
- ML is a small part of a broad portfolio (education, plant science, Africa, arts)
- Support concentrated at UCL
- Website
- gatsby.org.uk
- Kind
- Foundation / funder
- Location
- United Kingdom
- Founded
- 1967
- Funding scale
- Over £1.7bn committed since 1967 across all programmes
Focus areas
- Theory
- Probabilistic ML
- ML for science
How to engage
- Commissioned programmes (no open AI call stated)
- Recruiting via the Gatsby Unit PhD programme
Sources · verified November 2026
Anchor: #eu-gatsby-charitable-foundation
Hector Stiftung (Hector Foundation)
Researchers and institutions seeking large private endowment-style support for European AI research, especially in the Tübingen/Cyber Valley ecosystem.
High confidenceDetailsClose
Private foundations of Josephine and Hans-Werner Hector (est. 1995; Hector Stiftung II 2008) funding science, education and medical research; it committed €100m to found the ELLIS Institute Tübingen in 2023.
Strengths
- Single largest private commitment to a European ML research institute (€100m to ELLIS Tübingen)
- Funds endowed PI positions designed to compete with industry labs
- Long track record in science and medical-research philanthropy since 1995
- Close ties to Cyber Valley and Tübingen's AI ecosystem
Watch-outs
- AI giving is concentrated on one institute rather than open calls
- Public site gives no overall budget or AI-specific application route
- Regional focus (Baden-Württemberg) for most programmes
- Website
- hector-stiftung.com
- Kind
- Foundation / funder
- Location
- Weinheim, Germany
- Founded
- 1995
- Funding scale
- €100m commitment to the ELLIS Institute Tübingen (2023)
Flagship outputs
Focus areas
- Trustworthy / safe AI
- ML for science
How to engage
- Project funding application (Projektförderung)
- Hector Endowed Fellow positions (via ELLIS Institute Tübingen)
- Strategic philanthropy
Sources · verified November 2026
Anchor: #eu-hector-foundation
Helmholtz AI
Scientists and companies seeking ML collaborations on large scientific data (climate, health, energy, materials) with Helmholtz centres; ML researchers looking for science-facing research-lab roles.
Medium confidenceDetailsClose
Applied-AI platform of Germany's Helmholtz Association (launched 2019 as HAICU) providing AI research labs, consultant teams, GPU access (HAICORE) and project funding across Helmholtz centres; also runs the Helmholtz Foundation Model Initiative.
Strengths
- Rated 'excellent' (grade A) in the Oct 2022 external evaluation; consultant teams drew the most praise
- Access to unique scientific datasets and HPC across Helmholtz centres (e.g., JSC, KIT; HAICORE GPUs)
- Funded foundation-model programme for science domains (7 HFMI pilot projects)
- 11 AI research labs listed (e.g., teams Akata, Fortuin, Kilbertus)
Watch-outs
- Funding and services are mainly for Helmholtz researchers (project calls need at least two Helmholtz centres); externals join as partners
- No public overall budget figure on the platform's site
- Applied-science focus: less core ML-method output than university ML groups
Leading researchers
- Scientific Director, Helmholtz AIComputational biology / single-cell ML; Professor at TUM and head of Institute of Computational Biology, Helmholtz Munich
- Managing Director, Helmholtz AIRuns operations, calls and the platform's consultant model
- Head of AI Consultant team, Jülich Supercomputing CentreLarge-scale model training on supercomputers; HAICORE contact
- Website
- helmholtz.ai
- Kind
- Research institute
- Institution
- Helmholtz Association (central unit at Helmholtz Munich)
- Location
- Munich (central unit) and local units at Helmholtz centres, Germany
- Founded
- 2019
- Funding scale
- Foundation Model Initiative: €11m for pilots + €12m infrastructure (from 2024); project calls fund up to €250k per project, matched by centres (≥€500k total)
Flagship outputs
Focus areas
- ML for science
- Computer vision
- Efficient ML
- Probabilistic ML
How to engage
- AI consulting / vouchers (Helmholtz researchers)
- Helmholtz AI Associate membership
- Industry or academic partner on Helmholtz AI projects
- Hiring (AI research labs)
- HAICON conference
Sources · verified November 2026
Anchor: #eu-helmholtz-ai
IDEAS Research Institute (Instytut Badawczy IDEAS)
Companies and public bodies seeking Polish AI R&D partners (incl. defence/security and medical imaging); students and young researchers via hackathons, summer schools and ELLIS Unit Warsaw.
Medium confidenceDetailsClose
State AI research institute created by Council of Ministers regulation (25 Feb 2025) and directed by Piotr Sankowski, founder of the earlier IDEAS NCBR; it lists 13 research areas from vision and medical imaging to NLP and cybersecurity, and is the main host of ELLIS Unit Warsaw (directed by its R&D director Tomasz Trzciński; co-hosted since Dec 2025 by the University of Warsaw, listed separately as warsaw-uw-ellis-warsaw, and WUT).
Strengths
- Led by the founder of IDEAS NCBR, which built Poland's leading AI research team network
- Main host of ELLIS Unit Warsaw (with UW and WUT as co-hosts) and runs European events (IDEASHACK, MLSS Reliability & Safety 2026)
- Broad portfolio: 13 research areas listed, incl. machine vision, medical imaging, NLP, GenAI and cybersecurity
Watch-outs
- Young (2025) institution created after the 2024 leadership change at IDEAS NCBR; track record as a separate entity is short
- Modest public start-up budget (PLN 20m) and reliance on corporate sponsors such as ORLEN
- Defence and public-administration mandate may constrain open research or foreign partnerships
Leading researchers
- Director, IDEAS Research InstituteFounded and led IDEAS NCBR 2021–2024; named Director of IDEAS Research Institute in ITwiz (May 2026)
- R&D Director, IDEAS Research Institute; director of ELLIS Unit Warsaw (IDEAS/WUT)Quoted as R&D director on IDEASHACK 2026
- Website
- ideas.edu.pl
- Kind
- Research institute
- Institution
- Polish state research institute supervised by the Ministry of Digital Affairs and the Ministry of National Defence
- Location
- Warsaw, Poland
- Founded
- 2025
- Funding scale
- PLN 20m from the Ministry of Digital Affairs for set-up and launch (announced Feb 2025); ORLEN financial sponsorship from 2026 (amount undisclosed)
Flagship outputs
Focus areas
- Computer vision
- NLP / LLMs
- Trustworthy / safe AI
- Efficient ML
How to engage
- Industry R&D partnership (e.g., ORLEN, strategic partner and sponsor since May 2026)
- Hackathons and summer schools
- ELLIS Unit Warsaw events
- Hiring (research groups)
Sources · verified November 2026
Anchor: #eu-ideas-research-institute
IDSIA – Dalle Molle Institute for Artificial Intelligence (USI-SUPSI)
Swiss and international firms seeking applied ML, Bayesian-network/causal modelling, optimisation or robotics projects co-funded by Innosuisse, and partners in finance, pharma and payments (UBS, Novartis, Mastercard listed).
Medium confidenceDetailsClose
Founded in 1988 by Angelo Dalle Molle and public since 2000, IDSIA is a joint institute of USI's Faculty of Informatics and SUPSI with about 130 staff (2023). It is headed by Andrea Emilio Rizzoli with three scientific directors (Jürgen Schmidhuber, Marco Zaffalon, Cesare Alippi), is known for early neural-network work under Schmidhuber, and today spans ML, robotics, probabilistic reasoning and NLP.
Strengths
- Long track record (since 1988) and historic role in neural-network and deep-learning research
- Dual university / applied-sciences affiliation (USI + SUPSI), described by USI as unique in Switzerland, suits applied projects
- Named industry partners in finance, pharma and payments (UBS, Novartis, Mastercard) plus Innosuisse funding
- Distinctive probabilistic-graphical-model and causal-discovery expertise (bnlearn, imprecise probability)
Watch-outs
- Low ML-core venue output (USI #102 in Europe, CSRankings 2021–2025) compared with its reputation
- Scientific director Schmidhuber concurrently directs the KAUST AI Initiative in Saudi Arabia
- Not listed as part of an ELLIS unit; institute homepage shows placeholder statistics rather than real figures
Leading researchers
- One of IDSIA's three scientific directors (with Marco Zaffalon and Cesare Alippi); Professor of AI, USI; Professor, SUPSI; also Director of the KAUST AI Initiative (Saudi Arabia)Pioneer of recurrent networks and deep learning; co-founder of several AI companies
- Marco ScutariTop 2%Senior Researcher in Bayesian Networks and Graphical Models, IDSIAlisted on the topcitedscientists.com Switzerland ICT page (IDSIA; Swiss ICT #26). Author of bnlearn
- Luca Maria GambardellaTop 2%Full Professor of AI, USI Faculty of Informatics; IDSIA member and its director 1995–2020; since late 2025 also in a private-sector rolelisted on the topcitedscientists.com Università della Svizzera italiana page (AI & Image Processing). Ant-colony optimisation and robotics
- Website
- idsia.usi-supsi.ch
- Kind
- Research institute
- Institution
- Università della Svizzera italiana (USI) and SUPSI
- Location
- Lugano, Switzerland
- Founded
- 1988
- CSRankings signal
- Università della Svizzera italiana: #102 in Europe for ML core (2.1 adj.), #55 for vision, 2021–2025; SUPSI staff are not covered by CSRankings (CSRankings data)
- Funding scale
- Competitive funding from Innosuisse, the Swiss National Science Foundation and ERC projects (totals not stated); ~130 staff
Flagship outputs
Focus areas
- Probabilistic ML
- Causality
- Robotics learning
- NLP / LLMs
- Theory
How to engage
- Innosuisse co-funded industry projects
- PhD (USI Faculty of Informatics)
- Contract research
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-idsia-usi-supsi
Imperial College London AI (I-X and Department of Computing)
Industry partners in vision/SLAM, neural-network verification, safe autonomy, affective computing and AI-for-science seeking co-location at White City, Schmidt AI-in-Science postdocs, or MSc/PhD hires.
Medium confidenceDetailsClose
Imperial's AI strength spans Computing and Engineering. I-X, launched in Feb 2023 at White City, is the college's flagship AI initiative, co-locating 20+ academic AI staff and 100+ PhD students and postdocs from twenty departments with industry partners.
Strengths
- Balanced profile: top-10 in Europe for ML core, vision (#4) and classic AI (#3) in CSRankings 2021–2025
- Distinctive verification and safe-autonomy expertise (neural-network verification, argumentation)
- I-X co-locates academic AI staff, PhD students and industrial partners at White City (I-HUB, 84 Wood Lane)
- Large AI-in-science postdoc programme funded by Schmidt Futures
Watch-outs
- Not an ELLIS Unit site: the ELLIS Unit London page lists only UCL, although Imperial hosts individual ELLIS Fellows (e.g., Lucia Specia)
- ML-core output lower than Oxbridge despite many active faculty (33)
- I-X is a broad cross-faculty initiative; core ML groups remain spread across departments
Leading researchers
- Andrew DavisonTop 2%Professor, Department of Computingin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Imperial; AI & Image Processing; UK ICT #14). Visual SLAM and spatial AI for robots
- Stefanos ZafeiriouTop 2%Professorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Imperial; AI & Image Processing). Face and human analysis in vision; high Imperial CV output in CSRankings
- FacultyApproximate inference and generative models; top Imperial ML-core output 2021-25
- ProfessorComputational argumentation and explainable AI
- ProfessorVerification of neural networks and multi-agent systems
- Björn W. SchullerTop 2%Faculty, Department of Computingin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Imperial; AI & Image Processing (Imperial rank 6); UK ICT #5). Affective computing and audio/speech ML; ranked #5 in UK ICT on the 2026 list
- Website
- ix.imperial.ac.uk
- Kind
- University group / lab
- Institution
- Imperial College London
- Location
- London, United Kingdom
- Founded
- 2023
- CSRankings signal
- #10 in Europe for ML core (38.5 adj.), #4 for vision, #3 for AI (AAAI/IJCAI); #8 combined, 2021–2025 (CSRankings data)
- Funding scale
- Schmidt AI in Science fellowship: up to 20 postdocs/yr for up to 6 years
Flagship outputs
Focus areas
- Computer vision
- Trustworthy / safe AI
- Robotics learning
- Probabilistic ML
- NLP / LLMs
- ML for science
How to engage
- PhD / UKRI CDTs
- Industry partnership programme (co-location at I-X, White City)
- Spin-out / tech transfer
- Postdoctoral fellowships
Sources · verified November 2026
Anchor: #eu-imperial-i-x
Inria (French National Institute for Research in Digital Science and Technology)
Partners seeking open-source ML tooling (scikit-learn via Probabl), optimisation/theory and vision collaborators, joint Inria labs, and start-up creation or licensing through Inria's transfer programmes.
Medium confidenceDetailsClose
France's national digital-science institute (~2,600 staff in project-teams), co-leading the €73m PEPR AI programme. Its researchers anchor the ELLIS Units Paris, Grenoble and Côte d'Azur.
Strengths
- Origin and long-time steward of scikit-learn; spun out Probabl in 2023 to sustain it
- Deep bench in optimisation/theory (Bach) and vision (Schmid) underpinning ELLIS Unit Paris
- National scale (~2,600 staff) with centres across France and structured start-up programmes
- Co-leads the €73m PEPR AI programme on frugal, trustworthy and mathematical AI
Watch-outs
- Bibliometric rankings undercount Inria because staff often publish under university affiliations
- Large national bureaucracy; contracting often involves multiple co-supervising bodies
- Expertise dispersed across many project-teams and cities; needs a targeted entry point
Leading researchers
- Francis BachTop 2%Research Director; ELLIS Paris Programme Directorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under Université PSL; France ICT #5). Optimisation and kernel methods for ML
- Cordelia SchmidTop 2%Research Director; ELLIS Paris Programme Directorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (INRIA; AI & Image Processing; France ICT #7). Computer vision and video understanding
- Michael I. JordanTop 2%Research scientist, Inria Paris; Pehong Chen Distinguished Professor, UC Berkeleyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under UC Berkeley; AI & Image Processing). Statistics, ML and economics; top CSRankings ML-core output among Inria-listed faculty
- Research Director (scikit-learn); Probabl co-founderscikit-learn co-creator; ML for health/tabular data
- Website
- inria.fr
- Kind
- Research institute
- Location
- Paris (HQ) and centres across France, France
- Founded
- 1967
- CSRankings signal
- INRIA: #15 in Europe for ML core (27.2 adj.) from only 4 listed researchers, 2021–2025 — CSRankings lists few Inria staff, so France is undercounted
- Funding scale
- Co-leads PEPR AI: €73m over six years (France 2030)
Flagship outputs
Focus areas
- Theory
- Computer vision
- Efficient ML
- Trustworthy / safe AI
- ML for science
- Probabilistic ML
How to engage
- PhD / ELLIS PhD programme
- Industry partnership / joint labs
- Spin-out / tech transfer
- Licensing
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-inria
INSAIT – Institute for Computer Science, Artificial Intelligence and Technology
Companies seeking CEE-based AI talent, sovereign/low-resource-language LLMs (Bulgarian, Ukrainian) or computer-vision collaboration with strong Swiss links; sponsorship is the main industry route.
Medium confidenceDetailsClose
Research institute launched April 2022 in Sofia, founded by ETH professor Martin Vechev in partnership with ETH Zurich and EPFL; it hosts the ELLIS Unit Sofia and has released national open LLMs (BgGPT, MamayLM).
Strengths
- Rapid rise: #5 in Europe for computer-vision output within a few years of launch
- Built-in links to ETH Zurich and EPFL; advisory board of senior international researchers
- Track record of national open LLMs for under-served languages (Bulgarian, Ukrainian)
- Backed by major tech donors (logos incl. Google, DeepMind, AWS, VMware)
Watch-outs
- Output concentrated in one senior vision PI (most of INSAIT's CV count)
- Young institute (2022); long-term funding structure not public on sources reviewed
- Smaller local industry ecosystem than Western European hubs
Leading researchers
- Martin VechevTop 2%Founder & Scientific Directorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (ETH Zürich; Swiss ICT #65). Reliable/secure AI; also ETH Zurich professor
- Luc Van GoolTop 2%Professor (INSAIT); also heads computer-vision groups at ETH Zurich and KU Leuvenin the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under Sofia University St. Kliment Ohridski; Bulgaria #3). Vision pioneer; ERC Advanced Grant; drives INSAIT's #5 European CV ranking
- Researcher3D vision and geometric learning
- ResearcherTrustworthy and collaborative ML
- Website
- insait.ai
- Kind
- Research institute
- Institution
- Sofia University (in partnership with ETH Zurich and EPFL)
- Location
- Sofia, Bulgaria
- Founded
- 2022
- CSRankings signal
- #36 in Europe for ML core (9.8 adj.) but #5 for vision (CVPR/ECCV/ICCV) and #18 combined, 2021–2025 (CSRankings data)
Focus areas
- Computer vision
- NLP / LLMs
- Robotics learning
- Trustworthy / safe AI
How to engage
- PhD programme
- Industry partnership / sponsorship
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-insait
Institut Polytechnique de Paris & Hi! PARIS (with HEC Paris)
Companies wanting statistics-heavy ML, optimal transport, graph/LLM methods or AI-for-business partners, plus engineering and business-school graduate hiring through programmes, career fairs and a research engineering team.
Medium confidenceDetailsClose
Hi! PARIS is the AI and data-science centre created by IP Paris and HEC Paris (Inria joined later; executive director Vincent Rapp since Mar 2025). In May 2024 its consortium won €70m over five years as the France 2030 IA Cluster 'Hi! PARIS Cluster 2030' (ANR coordinator: Gaël Richard, Télécom Paris; Eric Moulines was announced as project head). Paris's other IA Cluster, PSL-led PR[AI]RIE-PSAI, is listed separately; both are participating institutions of ELLIS Unit Paris, which is directed from ENS (see psl-ens-prairie-psai).
Strengths
- €70m France 2030 IA Cluster award (May 2024) on top of an established centre with 22 chairs and 15 ERC grants
- Strong statistics and optimal-transport bench (Moulines, Flamary, Durmus) with widely used open-source POT
- Engineering + business-school pairing (IP Paris and HEC) suits applied AI-for-business partnerships
- Large training pipeline: 40 AI-related programmes and 3,000+ students involved since 2021
Watch-outs
- Invisible in CSRankings (no IP Paris faculty entered), so output cannot be benchmarked there
- Five schools plus HEC, Inria, CNRS and others in the consortium; partners must identify the right lab
- Corporate donors and the centre's founding date are not named on its homepage; the cluster head announced in 2024 (Moulines) now also holds a dean post at MBZUAI
Leading researchers
- Eric MoulinesTop 2%Professor of statistics, École polytechnique (CMAP); since 2025 also Dean of the Computing and Mathematical Sciences Division, MBZUAI (Abu Dhabi); announced as head of Hi! PARIS Cluster 2030 (May 2024); ELLIS Fellowlisted on the topcitedscientists.com École Polytechnique page (Networking & Telecommunications). Statistical learning and Monte Carlo methods
- Rémi FlamaryTop 2%Professor, École polytechnique (CMAP); co-director of the LLMs, Graphs and Applications MSc track; ELLIS Scholarlisted on the topcitedscientists.com École Polytechnique page (AI & Image Processing). Co-creator of the POT library
- Michalis VazirgiannisTop 2%Distinguished Professor, École polytechnique; co-director of the LLMs, Graphs and Applications MSc tracklisted on the topcitedscientists.com École Polytechnique page (AI & Image Processing). Graph ML and NLP
- Website
- hi-paris.fr
- Kind
- University group / lab
- Institution
- Institut Polytechnique de Paris (École polytechnique, ENSTA, ENSAE, Télécom Paris, Télécom SudParis); HEC Paris
- Location
- Palaiseau (Paris-Saclay), France
- CSRankings signal
- Not ranked: CSRankings lists IP Paris as an institution but has no faculty entered for it, so output is uncounted (CSRankings data, 2026-11)
- Funding scale
- €70m over five years from France 2030 IA Cluster (announced May 2024; ANR project Dec 2024–Dec 2029); centre reports 22 chair holders, 50 PhD students and 75 postdoc fellowships
Flagship outputs
Focus areas
- Probabilistic ML
- Theory
- NLP / LLMs
- Computer vision
How to engage
- Corporate donor / membership (5 corporate donors listed)
- PhD and postdoc fellowships / ELLIS PhD programme
- Graduate hiring (Hi! PARIS Career Fair, Data Boot Camp, Hi!ckathon)
- Start-up incubation via the Cluster 2030 programme
Sources · verified November 2026
Anchor: #eu-ip-paris-hi-paris
Instituto Superior Técnico AI (ELLIS Unit Lisbon: IST, INESC-ID, IT, ISR)
Partners in multilingual NLP, machine translation and evaluation, sparse/structured deep learning, social and service robotics, and companies wanting Portugal-based PhD talent or EU-funded LLM collaborations.
Medium confidenceDetailsClose
Lisbon's AI research centres on Instituto Superior Técnico and its associated institutes (INESC-ID, Instituto de Telecomunicações, ISR-Lisboa), joined since May 2020 as ELLIS Unit Lisbon; it is Portugal's strongest ML/NLP cluster and co-builds the open EuroLLM models with Unbabel.
Strengths
- Portugal's leading ML/NLP cluster: #13 AI and #21 NLP in Europe (CSRankings 2021–2025)
- Builds open European LLMs (EuroLLM, Tower) with industry partner Unbabel and EuroHPC compute
- Large ELLIS unit (24 listed people incl. 8 Fellows) spanning NLP, robotics, speech and logic/SAT
- Large engineering school base: 12,017 students and 1,056 faculty/researchers at IST
Watch-outs
- Modest ML-core venue output (#60 in Europe) compared with its NLP/AI strength
- Research spread across several associated institutes (INESC-ID, IT, ISR), complicating a single point of contact
- Funding scale and lab headcounts not stated in the sources reviewed
Leading researchers
- Mário A. T. FigueiredoTop 2%Professor, IST; ELLIS Unit Lisbon directornamed on his topcitedscientists.com profile page (IST; AI & Image Processing, subfield rank #1,023); not shown on the Universidade de Lisboa institution page. Signal processing and sparse learning
- André F. T. MartinsTop 2%Associate Professor, IST; leads SARDINE Lab (IT); VP of AI Research, Unbabel; ELLIS Fellow and NLP programme co-directorlisted on the topcitedscientists.com Universidade de Lisboa page (AI & Image Processing). EuroLLM, Tower, xCOMET; ERC StG and CoG
- Website
- sites.google.com
- Kind
- University group / lab
- Institution
- Universidade de Lisboa – Instituto Superior Técnico
- Location
- Lisbon, Portugal
- Founded
- 2020
- CSRankings signal
- Universidade de Lisboa: #60 in Europe for ML core (5.3 adj.), #13 for AI (AAAI/IJCAI), #21 for NLP, #20 for robotics, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- NLP / LLMs
- Theory
- Robotics learning
- Trustworthy / safe AI
How to engage
- PhD / ELLIS PhD programme
- Industry partnership (e.g. Unbabel co-appointments and joint LLM projects)
- EU Horizon / EuroHPC consortium partnering
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-ist-lisbon-ellis-lisbon
Integreat – Norwegian Centre for Knowledge-driven Machine Learning (University of Oslo & UiT)
Partners in statistically grounded, knowledge-driven or explainable ML; medical imaging, remote sensing and energy-sector image analysis; Norwegian-language models; and PhD recruitment in Norway.
Medium confidenceDetailsClose
Research Council of Norway Centre of Excellence (launched July 2023, funded to 2033) that blends statistics, logic and ML; its partners also run the UiT-led Visual Intelligence innovation centre and Oslo's Norwegian-language LLM work. Integreat co-director Robert Jenssen (UiT) co-directs ELLIS Unit Norway (est. 1 Oct 2026, seven Norwegian institutions).
Strengths
- Ten-year Centre of Excellence status (2023–2033) gives unusually stable funding for knowledge-driven ML
- Distinctive statistics + logic + ML mix with ~30 researchers and ~80 PhD students
- Industry-facing imaging work via the Visual Intelligence centre (Equinor, Aker BP, GE Healthcare, KSAT)
- Open Norwegian LLMs and encoders (NorBERT, NorMistral) from the Oslo Language Technology Group
Watch-outs
- Small ML-core venue footprint: UiT #58 in Europe; University of Oslo not covered by CSRankings
- Modest budget (~€16m over ten years) and UiO notes funding cuts are reducing planned activities
- Activity split between Oslo and Tromsø and several centres, so partners must pick an entry point
Leading researchers
- Professor of statistics and data science, University of Oslo; Integreat DirectorCo-obtained Integreat's Centre of Excellence funding in 2022; chair of the Abel Prize board
- Professor, Department of Biostatistics (Institute of Basic Medical Sciences), University of Oslo; founding director and co-director of IntegreatBayesian and computational statistics; leads the BigInsight innovation centre; ELLIS member
- Professor, UiT Machine Learning Group; Director, Visual Intelligence (SFI); Integreat co-director; ELLIS Unit Norway co-director (with Jukka Corander and Michael Kampffmeyer)Deep learning for medical and remote-sensing images with industry partners
- Website
- integreat.no
- Kind
- University group / lab
- Institution
- University of Oslo (host); UiT The Arctic University of Norway; Norwegian Computing Center
- Location
- Oslo / Tromsø, Norway
- Founded
- 2023
- CSRankings signal
- UiT: #58 in Europe for ML core (5.6 adj.), 2021–2025; University of Oslo is not listed as an institution in CSRankings (CSRankings data)
- Funding scale
- Integreat: ~€16m over ten years (2023–2033) from the Research Council of Norway, reduced by funding cuts per UiO
Flagship outputs
Focus areas
- Probabilistic ML
- Trustworthy / safe AI
- ML for science
- NLP / LLMs
- Computer vision
How to engage
- PhD (Integreat reports ~80 PhD students) / ELLIS PhD programme
- Industry consortium membership (Visual Intelligence SFI, BigInsight)
- Visiting / guest researcher
- Master's student projects
Sources · verified November 2026
Anchor: #eu-integreat-oslo-uit
ISTA Machine Learning (ELLIS Unit Vienna)
LLM-inference, compression and efficient-training teams seeking collaborators or licences; causal-representation-learning partners; investors tracking ISTA spin-outs via xista.
High confidenceDetailsClose
Graduate-research institute near Vienna whose small ML faculty (Alistarh, Lampert, Locatello, Mondelli) punches above its weight; it hosts the ELLIS Unit Vienna (est. Oct 2019).
Strengths
- Very high output per researcher: #11 in Europe for ML core from just 8 faculty (CSRankings 2021–2025)
- World-leading LLM compression and quantisation research (GPTQ, SparseGPT, Marlin kernels); GPTQ models downloaded millions of times
- Proven exit: Neural Magic, ISTA's first start-up collaboration with a major exit (to Red Hat)
- Dedicated tech-transfer arm (xista innovation) for licensing and spin-offs
Watch-outs
- Small ML group heavily dependent on a few PIs; key GPTQ/SparseGPT author Elias Frantar has left for OpenAI
- Limited NLP and applied-vision breadth compared with large universities
- Location outside central Vienna; ELLIS Unit Vienna is ISTA-only
Leading researchers
- Professor (since 2022), DASLab; ML Research Lead at Neural Magic (Red Hat)LLM compression (GPTQ, SparseGPT); led ML research at Neural Magic
- Christoph H. LampertTop 2%Professor; ELLIS Unit Vienna directorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (ISTA; AI & Image Processing). Trustworthy ML and computer vision
- Assistant ProfessorCausal and object-centric representation learning
- Assistant ProfessorTheory of deep learning and information theory
- Website
- ist.ac.at
- Kind
- Research institute
- Institution
- Institute of Science and Technology Austria (ISTA)
- Location
- Klosterneuburg (Vienna), Austria
- Founded
- 2009
- CSRankings signal
- #11 in Europe for ML core (32.6 adj.) with only 8 ML-active faculty — highest per-faculty output among top-15, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Efficient ML
- Causality
- Theory
- Trustworthy / safe AI
- Computer vision
How to engage
- PhD (ISTA Graduate School)
- ELLIS PhD programme
- Spin-out / tech transfer (xista innovation)
- Industry collaboration / licensing
Sources · verified November 2026
Anchor: #eu-ista-machine-learning
JKU Institute for Machine Learning & ELLIS Unit Linz
Industrial partners interested in recurrent/sequence architectures (xLSTM), deep learning for science and time series, and graduates of JKU's dedicated AI degrees; engage via the LIT AI Lab.
Medium confidenceDetailsClose
Institute headed by LSTM co-inventor Sepp Hochreiter, hosting the ELLIS Unit Linz at the LIT AI Lab (budget ~€25m over five years) and JKU's AI bachelor's and master's programmes.
Strengths
- Origin of LSTM and its 2024 successor xLSTM, a European alternative sequence architecture
- ELLIS Unit Linz with ~€25m five-year budget including industrial co-funding
- Dedicated AI bachelor's and master's programmes feed a local talent pipeline
- Hochreiter holds a guest seat on the ELLIS board, giving network visibility
Watch-outs
- Extreme concentration on one PI; CSRankings lists a single ML-core author at JKU
- Low ML-core venue output in CSRankings (#56 in Europe) relative to reputation
- Smaller ecosystem in Linz than in major capitals
Leading researchers
- Sepp HochreiterTop 2%Head of Institute; ELLIS Linz founding leadin the V9 (Aug 2026) list per the topcitedscientists.com mirror (JKU Linz; AI & Image Processing; global rank #117). Co-inventor of LSTM; xLSTM; on ELLIS board (guest)
- Gerhard WidmerTop 2%Professor, Institute of Computational Perceptionin the V9 (Aug 2026) list per the topcitedscientists.com mirror (JKU Linz; AI & Image Processing). AI for music and audio
- Website
- jku.at
- Kind
- University group / lab
- Institution
- Johannes Kepler University (JKU) Linz
- Location
- Linz, Austria
- CSRankings signal
- JKU Linz: #56 in Europe for ML core (6.0 adj.), essentially one CSRankings-listed ML author, 2021–2025 (CSRankings data)
- Funding scale
- ELLIS Unit Linz: ~€25.0m over five years (university and industrial partners)
Flagship outputs
Focus areas
- Efficient ML
- ML for science
- RL
- NLP / LLMs
- Computer vision
How to engage
- PhD / ELLIS PhD programme
- Industry partnership (LIT AI Lab)
- AI BSc/MSc graduate hiring
- Spin-out / tech transfer
Sources · verified November 2026
Anchor: #eu-jku-institute-for-machine-learning
Knut and Alice Wallenberg Foundation (KAW)
Swedish-based PIs (and their international co-PIs) seeking AI-for-science funding, PoC money for AI spin-outs, or GPU compute; life-science ML groups wanting to join DDLS.
High confidenceDetailsClose
One of Europe's largest private research foundations (founded 1917; over SEK 2.5bn in grants in 2025); beyond WASP it funds Data-Driven Life Science (SEK 3.7bn), the Berzelius AI supercomputer, AI proof-of-concept grants and a free AI-consultancy grant for its grantees.
Strengths
- Scale and patience: century-old endowment funding multi-billion SEK, multi-year programmes (WASP, DDLS)
- Concrete AI instruments: PoC grants (SEK 1–4m), free AI consultancy hours, Berzelius GPU upgrades
- Links AI with domain science (life science, materials, quantum) rather than funding AI in isolation
Watch-outs
- Most calls require a PI employed at a Swedish university; international co-PIs cannot receive funds
- Several AI instruments (consultancy, PoC) are restricted to existing Wallenberg grantees or programmes
- Leadership transition at the turn of 2026/27 may shift priorities
Leading researchers
- Chair, Knut and Alice Wallenberg FoundationQuoted on the 2024 AI and life science initiative
- Executive Director (stepping down at the turn of 2026/27)Eight years at the foundation, almost three as Executive Director; led the 2024 DDLS and AI expansions
- Incoming Executive Director (announced June 2026)Takes over at the turn of the year; Anders Ynnerman and Siv Andersson become Deputy Executive Directors
- Website
- kaw.wallenberg.org
- Kind
- Foundation / funder
- Location
- Stockholm, Sweden
- Founded
- 1917
- Funding scale
- Over SEK 2.5bn in grants in 2025 (KAW); almost SEK 42bn awarded since 1917; DDLS SEK 3.7bn total; WASP programme listed as a separate entry
Flagship outputs
- Data-Driven Life Science (DDLS): SEK 3.7bn total incl. SEK 270m 'Alpha Cell' AI cell-model pilot (2024)
- AI for Science via WASP (SEK 70m) and SEK 125m Berzelius AI supercomputer upgrade (2024)
- KAW AI Consultancy Grant: up to 700 free consultant hours for Wallenberg-funded PIs (current round 20 Nov–20 Dec 2026)
- KAW Proof of Concept Grants in AI, Quantum or ICT (SEK 1–4m, up to 2 years; yearly call, 2026 round closed 25 Mar 2026)
Focus areas
- ML for science
- Efficient ML
- Robotics learning
How to engage
- Grants / calls (PoC grants in AI; project grants; fellowships)
- AI Consultancy Grant (existing grantees)
- Compute via Berzelius (Swedish academia)
- Join strategic programmes (WASP, DDLS, WACQT)
Sources · verified November 2026
Anchor: #eu-knut-alice-wallenberg-foundation
KTH Royal Institute of Technology AI (RPL, Speech Music & Hearing; ELLIS Unit Sweden)
Robotics, manipulation and autonomous-systems partners; conversational and speech AI (TTS, turn-taking, social robots); graph mining; industrial PhDs co-funded through WASP.
Medium confidenceDetailsClose
Sweden's largest technical university hosts the Robotics, Perception and Learning (RPL) department, a strong speech and conversational-AI group and graph/data-mining research; it leads ELLIS Unit Sweden (est. April 2026) and is a core WASP partner.
Strengths
- Leads ELLIS Unit Sweden (April 2026) spanning KTH, Uppsala, Linköping and Chalmers
- Established robotics and manipulation school (Kragic) with WASP-funded industrial PhD links
- Distinctive speech and conversational-AI strength with an industry spin-out (Furhat Robotics) and open TTS models
- 19 KTH authors in the visible top 100 of the 2026 Stanford–Elsevier Sweden ICT list (vs 15 Chalmers, 7 Linköping), though many are in telecoms rather than ML
Watch-outs
- Modest ML-core output (#44 in Europe, CSRankings 2021–2025) and weak NLP-venue output
- AI expertise spread across several departments and schools; no single AI institute entry point
- Department headcounts and dedicated AI funding are not stated on the pages reviewed
Leading researchers
- Professor of Computer Science (RPL); Director, Centre for Autonomous Systems; Wallenberg Scholar; ELLIS Fellow and director of ELLIS Unit Sweden (est. 1 Apr 2026)Robotic grasping, manipulation and vision; ERC Starting Grant 2012; member of the Royal Swedish Academy of Sciences
- Aristides GionisTop 2%Professor, Theoretical Computer Science; WASP professorlisted on the topcitedscientists.com Sweden ICT page (KTH). Graph mining; ERC project REBOUND on online polarisation
- Gabriel SkantzeTop 2%Professor in Speech Communication and Technology; part-time Chief Scientist, Furhat Roboticslisted on the topcitedscientists.com Sweden ICT page (KTH). Conversational systems; co-founded Furhat Robotics (2014)
- Website
- kth.se
- Kind
- University group / lab
- Institution
- KTH Royal Institute of Technology
- Location
- Stockholm, Sweden
- Founded
- 1827
- CSRankings signal
- #44 in Europe for ML core (8.1 adj., 10 faculty), #23 for robotics (ICRA/IROS/RSS), 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Robotics learning
- Computer vision
- NLP / LLMs
- Probabilistic ML
How to engage
- PhD / ELLIS PhD programme
- Industrial PhD via WASP
- Spin-out / tech transfer (e.g. Furhat Robotics)
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-kth-ai-rpl
KU Leuven AI (DTAI, PSI & ELLIS Unit Leuven)
Partners seeking neuro-symbolic/probabilistic-logic AI, kernel methods, continual-learning vision or energy-efficient AI hardware collaboration, plus Belgian spin-out routes via KU Leuven R&D.
Medium confidenceDetailsClose
KU Leuven's AI groups span the Computer Science (declarative AI, probabilistic logic) and Electrical Engineering (vision, kernel methods, AI hardware) departments; the ELLIS Unit Leuven was established in January 2019.
Strengths
- Long-standing strength in probabilistic logic and statistical relational learning
- Rare combination of ML with AI-hardware design (Verhelst) in one ELLIS unit
- Six ELLIS Fellows across CS and EE departments
- Stronger classic-AI venue output (#21) than ML-core (#40) suggests reasoning/KR depth
Watch-outs
- Modest ML-core output (#40 in Europe, CSRankings 2021–2025)
- AI activity split across CS and EE departments; the Leuven.AI site (ai.kuleuven.be) blocks automated access, so details were checked via ELLIS pages
- Funding scale and institute-level structure not documented in sources reviewed
Leading researchers
- Professor; ELLIS Unit Leuven directorStructured prediction and learning theory; top Leuven ML-core output
- Full research professor; ELLIS FellowProbabilistic logic learning; ERC Advanced Grant (SYNTH)
- Professor; ELLIS FellowLanguage understanding and multimodal NLP
- Tinne TuytelaarsTop 2%Professor; ELLIS Fellowin the V9 (Aug 2026) list per the topcitedscientists.com mirror (KU Leuven; AI & Image Processing). Computer vision and continual learning
- Professor; ELLIS FellowEnergy-efficient AI hardware
- Johan SuykensTop 2%Faculty (ESAT)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (KU Leuven; AI & Image Processing (KU Leuven rank 13)). Kernel methods / least-squares SVMs; KU Leuven's highest-ranked AI & Image Processing author on the 2026 list
- Website
- ellis.eu
- Kind
- University group / lab
- Institution
- KU Leuven
- Location
- Leuven, Belgium
- Founded
- 2019
- CSRankings signal
- #40 in Europe for ML core (8.7 adj.), #21 for AI (AAAI/IJCAI), 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Probabilistic ML
- Trustworthy / safe AI
- Computer vision
- Efficient ML
- NLP / LLMs
How to engage
- PhD / ELLIS PhD programme
- Industry partnership programme
- Spin-out / tech transfer (KU Leuven R&D)
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-ku-leuven-ai
Kyutai
Teams building voice/speech products who want open speech models and code (or a commercial route via Gradium), and researchers seeking an open-science lab in Paris.
Medium confidenceDetailsClose
Paris-based non-profit open AI research lab launched in Nov 2023 and funded by Iliad Group, CMA CGM Group and Schmidt Sciences; known for speech-native models (Moshi, Hibiki, Pocket TTS). Its first spin-off, Gradium, commercialises its speech research.
Strengths
- Distinctive speech-native focus: Moshi described as the first speech-native dialogue system
- Steady stream of open releases (Pocket TTS, Muscriptor, dactory data codebase)
- Backed by large private funders (Iliad, CMA CGM, Schmidt Sciences)
- Proven commercial path: first spin-off Gradium (founded 2025 by Kyutai founding member Neil Zeghidour) raised a $70m seed round
Watch-outs
- Narrow focus on speech/audio relative to general-purpose labs
- Licensing varies by release; not all outputs state open licences on the homepage
- Small team (~28 people listed on the team page) and some founding researchers (Edouard Grave, Hervé Jégou) are now listed as alumni
Leading researchers
- Patrick PérezTop 2%Chief Executive Officerin the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under Kyutai; France ICT #24). Computer vision and ML; led the lab since its 2023 launch
- Head of ResearchAudio generation and speech-text modelling
- Neil ZeghidourTop 2%Audio Research Advisor (Kyutai); founder of spin-off Gradiumin the V9 (Aug 2026) list per the topcitedscientists.com mirror (named as top listed author for Gradium (global rank #119,690) in the France ICT institution table). Neural audio codecs and speech models; Kyutai founding member who founded Gradium in 2025 ($70m seed, Dec 2025)
- Website
- kyutai.org
- Kind
- Non-profit open lab
- Location
- Paris, France
- Founded
- 2023
- Funding scale
- Launched Nov 2023 with close to €300m committed (Xavier Niel €100m, Rodolphe Saadé/CMA CGM €100m, plus Eric Schmidt's foundation and other donors) and ~1,000 H100 GPUs at cost via Scaleway
Flagship outputs
Focus areas
- NLP / LLMs
- Efficient ML
How to engage
- Open-source models and code
- Research collaboration
- Hiring
Sources · verified November 2026
Anchor: #eu-kyutai
LAION e.V. (Large-scale Artificial Intelligence Open Network)
Open-source model builders and researchers needing open multimodal training data and reproducible pipelines; partners in open European foundation-model projects (e.g., MixtureVitae with OpenEuroLLM). Do legal/data-provenance review before commercial use.
Medium confidenceDetailsClose
German non-profit association (e.V.) building open, large-scale datasets, models and tools; creator of LAION-400M (2021) and LAION-5B (2022, used to train Stable Diffusion and Google's Imagen) and the cleaned Re-LAION-5B (2024).
Strengths
- Created the largest openly available image-text datasets (LAION-5B, 5bn+ pairs)
- Datasets underpinned widely used generative models (e.g., Stable Diffusion)
- Hamburg Regional Court (2024) dismissed a copyright suit, a landmark ruling on text-and-data-mining for AI training
- Active in European open-model efforts (MixtureVitae with OpenEuroLLM partners)
Watch-outs
- 2023 Stanford Internet Observatory report found suspected CSAM links; datasets withdrawn until Re-LAION-5B (2024)
- Volunteer/donation-funded structure; limited institutional capacity and SLAs
- Ongoing legal and reputational exposure around web-scraped training data
Leading researchers
- Co-founderInitiated LAION's open dataset efforts
- Co-founderScaling laws and open foundation-model research
- Website
- laion.ai
- Kind
- Non-profit open lab
- Location
- Germany
Flagship outputs
Focus areas
- Computer vision
- NLP / LLMs
- Efficient ML
How to engage
- Open-source contribution
- Donations
- Compute / research collaboration
- Joint EU projects
Sources · verified November 2026
Anchor: #eu-laion
Max Planck Institute for Intelligent Systems (MPI-IS)
Deep-tech founders and R&D teams needing causal ML, 3D human body models (SMPL licensing), haptics/robotics collaborators, or PhD talent via IMPRS-IS and Cyber Valley; licensing goes through Max Planck Innovation.
High confidenceDetailsClose
Basic-research institute (est. 2011) with ~600 staff across Tübingen (learning, perception, theory) and Stuttgart (physical intelligent systems). It initiated the Cyber Valley ecosystem and co-runs the IMPRS-IS doctoral school.
Strengths
- Home of foundational causal-inference ML (Schölkopf) and widely licensed 3D human models (SMPL)
- Anchor of Cyber Valley, whose partners include Amazon, Bosch and Fraunhofer, plus a start-up and investor network
- IMPRS-IS doctoral school (since 2017) shared with Stuttgart and Tübingen universities
- Unusual hardware + learning combination: robotics materials and haptics alongside ML theory
Watch-outs
- Max Planck focus on basic research; applied contract work is limited and IP terms go via Max Planck Innovation
- Department structure depends heavily on a handful of directors
- CSRankings reports all Max Planck institutes together, so MPI-IS-specific output is not separable
Leading researchers
- Bernhard SchölkopfTop 2%Director, Empirical Inferencein the V9 (Aug 2026) list per the topcitedscientists.com mirror (MPI for Intelligent Systems; AI & Image Processing; global rank #475). Kernel methods and causal ML; 2nd-highest European ML-core output in CSRankings 2021-25
- Michael J. BlackTop 2%Director, Perceiving Systemsin the V9 (Aug 2026) list per the topcitedscientists.com mirror (MPI for Intelligent Systems; AI & Image Processing). 3D human body modelling (SMPL); top European vision output in CSRankings
- Moritz HardtTop 2%Director, Social Foundations of Computationin the V9 (Aug 2026) list per the topcitedscientists.com mirror (MPI for Intelligent Systems; AI & Image Processing). Fairness, benchmarks and ML in society
- Director, Haptic IntelligenceHaptics and robot touch; former institute Managing Director
- Website
- is.mpg.de
- Kind
- Research institute
- Institution
- Max Planck Society
- Location
- Tübingen / Stuttgart, Germany
- Founded
- 2011
- CSRankings signal
- Max Planck Society (all MPIs incl. MPI-IS, MPI-INF, MPI-SWS): #6 in Europe for ML core (61.9 adj.), #2 for vision, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Causality
- Computer vision
- Robotics learning
- Probabilistic ML
- Trustworthy / safe AI
- Theory
How to engage
- PhD (IMPRS-IS)
- Cyber Valley industry partnership
- Spin-out / tech transfer
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-mpi-intelligent-systems
MIAI Cluster – Grenoble Alpes Interdisciplinary AI Institute (IA Cluster)
Researchers and PhD/postdoc candidates targeting Grenoble's AI chairs, companies seeking co-funded industrial or proof-of-concept chairs, and startups wanting the MIAI label and academic links.
Medium confidenceDetailsClose
France 2030 'IA Cluster' (labelled May 2024) succeeding the MIAI 3IA institute (2019); €70m state award, 12-partner consortium, ~600 AI researchers; focus on embedded, interactive and generative AI.
Strengths
- Joint second-largest IA Cluster award in France (€70m, level with Hi! PARIS), on top of a €212m track record since 2019
- Scale: ~600 AI researchers in 100+ labs (homepage); 1,850+ publications since 2019; 6 ERC grants among 18 awards (Feb 2025)
- Deep industry base: 100+ private partners and 150+ collaborative projects; industrial and PoC chair schemes
- Distinctive strength in embedded/frugal AI hardware, linked to CEA and Grenoble's microelectronics cluster
Watch-outs
- A funding umbrella across 12 partners rather than a single lab; chair-level budgets are not published
- Many chairs are applied (health, energy, heritage); core-ML depth is thinner than Paris or Germany's top hubs
- Calls are restricted to members of consortium partners; site partly French-only and several pages fail to load
Leading researchers
- Director, MIAI Cluster (appointment announced 16 Oct 2025)Succeeded founding director Éric Gaussier; previously innovation director at the Minalogic competitiveness cluster
- Scientific Director, MIAI Cluster (CNRS Research Director)Leads the MIAI scientific board
- Co-holder, THUNDER chair ('Theory and uncertainties in (Bayesian) deep learning', 2026–2029), with Pierre LatoucheChair in the 'machine learning and reasoning' core-AI category (May 2026–Sep 2029)
- Kind
- Network / society
- Institution
- Université Grenoble Alpes (coordinator) with Grenoble INP, CEA, CNRS, Inria, Inserm, CHU Grenoble Alpes, Université Clermont Auvergne, Université Savoie Mont Blanc and others
- Location
- Grenoble (with Clermont-Ferrand and Chambéry/Annecy), France
- Founded
- 2019
- Funding scale
- €70m France 2030 IA Cluster grant (5 years, from May 2024); €212m total budget since 2019 incl. partner funds; earlier 3IA grant €22m (2019–2026)
Flagship outputs
- MIAI chairs: 51 funded chairs, incl. 20 new chairs starting May 2026 (frugal AI, ML theory, AI4Science, health)
- Call for support from research engineers (renewed Feb 2026; GRICAD-hosted engineers to start or scale AI projects; open to UGA, UCA and USMB researchers)
- 2025 retrospective: 29 new chairs, FIDLE deep-learning training (21,500+ YouTube subscribers), MIAI Days
Focus areas
- Efficient ML
- ML for science
- Probabilistic ML
- Trustworthy / safe AI
- Robotics learning
How to engage
- Industrial chairs call (open 1 Mar 2025–31 Dec 2026; permanent researchers of consortium partners)
- Startup labelling (24 MIAI startups listed)
- PhD / postdoc and engineer positions via chair holders
- Academic–industry speed meetings and MIAI Days
Sources · verified November 2026
Anchor: #eu-miai-cluster-grenoble
MPI for Informatics & Saarland University (Saarland Informatics Campus)
Partners in 3D vision, digital humans, neural rendering and vision-language models; NLP/computational linguistics and AI planning collaborators; PhD hiring via IMPRS-TRUST and Saarland University.
High confidenceDetailsClose
The Max Planck Institute for Informatics (founded 1988, on the Saarland University campus since 1996) and Saarland University's computer science and language departments form the core of the Saarland Informatics Campus, home to ELLIS Unit Saarbrücken (est. Sept 2020).
Strengths
- Two MPI-INF departments with deep track records in human pose, 3D humans and neural rendering (Schiele, Theobalt)
- Industry-backed VIA centre (with Google, est. Nov 2022) added a vision-language-model research focus in Feb 2024
- Dense campus: ~1,100 researchers and 54 ERC grants across MPI-INF, MPI-SWS, CISPA, DFKI and Saarland University
- Strong computational linguistics at Saarland University (#10 NLP in Europe, CSRankings 2021–2025)
Watch-outs
- MPI-INF output is not separable in CSRankings; Saarland University's own ML-core output is mid-table (#39 in Europe)
- Max Planck basic-research mandate limits contract research; ML/vision strength concentrated in two director-led departments
- Financial terms of the Google partnership are not disclosed; campus spans several legal entities, so partners must pick the right one
Leading researchers
- Bernt SchieleTop 2%Director, Computer Vision and Machine Learning dept. (MPI-INF); Professor, Saarland University; ELLIS Unit Saarbrücken directorlisted on the topcitedscientists.com MPI for Informatics page (AI & Image Processing). Leads the MPI–Google vision-language-model focus
- Christian TheobaltTop 2%Director, Visual Computing and AI dept. (MPI-INF); Professor, Saarland University; founding director of VIA centrelisted on the topcitedscientists.com MPI for Informatics page (AI & Image Processing). Digital humans, neural rendering, motion capture
- Full Professor of Machine Learning, Saarland University; adjunct faculty MPI-SWS; ELLIS Fellow (ELLIS Unit Saarbrücken)Robust, interpretable and fair ML for consequential decisions
- Jörg HoffmannTop 2%Professor, Foundations of AI group, Saarland University; DFKI affiliate professorlisted on the topcitedscientists.com Universität des Saarlandes page (AI & Image Processing). AI planning; trust in learned action policies
- Kind
- Research institute
- Institution
- Max Planck Society; Saarland University
- Location
- Saarbrücken, Germany
- Founded
- 1988
- CSRankings signal
- Saarland University: #39 in Europe for ML core (8.7 adj.), #10 for NLP; MPI-INF is counted inside the Max Planck Society total (#6 ML core, #2 vision), 2021–2025 (CSRankings data)
- Funding scale
- Max Planck institutional funding; Google's additional VIA support amount not disclosed
Flagship outputs
Focus areas
- Computer vision
- NLP / LLMs
- Trustworthy / safe AI
- Theory
How to engage
- PhD (IMPRS-TRUST with Saarland University) / ELLIS PhD programme
- Industry partnership (VIA centre model, cf. Google)
- Spin-out / tech transfer (e.g. The Captury, markerless motion capture)
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-mpi-informatics-saarland-sic
Munich Center for Machine Learning (MCML)
Corporates and start-ups wanting vision/3D, graph-ML, AutoML or NLP collaborators, sponsored projects and a large Munich hiring pool; contract with LMU or TUM depending on the PI.
High confidenceDetailsClose
Joint LMU–TUM centre founded in 2018 and one of Germany's six permanently funded AI centres (€19.6m/yr from the federal government and Bavaria), with 50+ research groups.
Strengths
- Permanent institutional funding of €19.6m/yr from federal government and Bavaria
- TUM ranks #1 in Europe for computer-vision output and #5 for ML core (CSRankings 2021–2025)
- Scale: 50+ research groups across two universities
- Track record of high-impact open models (Stable Diffusion via LMU's CompVis lab)
Watch-outs
- Two-university structure means separate contracting/IP offices (LMU vs TUM)
- ML-core output is weighted to a few TUM PIs; LMU's strengths are more NLP/statistics
- Large centre with broad remit; smaller firms may find it hard to identify the right group
Leading researchers
- Daniel CremersTop 2%Director (TUM)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under Munich Center for Machine Learning; Germany ICT #25). 3D vision and SLAM; MCML director
- Stephan GünnemannTop 2%Professor (TUM)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under Munich Data Science Institute; Germany ICT #38). Graph ML and robustness; 5th-highest European ML-core output in CSRankings 2021-25
- Humboldt Professor (TUM)ML theory and graph learning; Humboldt Professorship in AI (2024); also linked to MIT (CSAIL homepage still active)
- Björn OmmerTop 2%Professor (LMU)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (LMU München; AI & Image Processing). Lab behind latent diffusion / Stable Diffusion (CompVis)
- Bernd BischlTop 2%Director (LMU)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under MCML; Germany ICT #68). AutoML and statistical learning; MCML director
- Website
- mcml.ai
- Kind
- Research institute
- Institution
- LMU Munich and Technical University of Munich (TUM)
- Location
- Munich, Germany
- Founded
- 2018
- CSRankings signal
- TUM: #5 in Europe for ML core (74.1 adj.), #1 for vision, #2 combined; LMU: #17 ML core, #7 NLP — ICML/NeurIPS/ICLR etc., 2021–2025 (CSRankings data)
- Funding scale
- €19.6m per year long-term federal/Bavarian funding (announced 2022)
Flagship outputs
Focus areas
- Computer vision
- NLP / LLMs
- Probabilistic ML
- ML for science
- Trustworthy / safe AI
How to engage
- PhD / ELLIS PhD programme
- Industry partnership programme
- ML consulting services
- Spin-out / tech transfer
Sources · verified November 2026
Anchor: #eu-munich-center-for-machine-learning
Novo Nordisk Foundation (AI programmes: DCAI / Gefion, Pioneer Centre for AI)
Life-science and AI researchers or start-ups in Denmark seeking compute (Gefion via DCAI) or open-competition grants at the bio-AI interface.
Medium confidenceDetailsClose
Large Danish enterprise foundation (DKK 11.8bn awarded in 2025) whose AI activity centres on the Danish Centre for AI Innovation, which owns and operates the Gefion AI supercomputer, and co-funding of the Pioneer Centre for AI.
Strengths
- Among Europe's largest research funders (DKK 11.8bn awarded in 2025)
- National AI compute through DCAI's Gefion supercomputer
- Co-funds the Pioneer Centre for AI (DKK 352.4m with four other foundations)
- Natural fit for AI-for-life-sciences projects
Watch-outs
- AI-specific budget lines and calls are not clearly broken out on public pages
- Funding priorities skew to health/life sciences; generic ML may fit less well
- Many programmes are Denmark-centred
- Website
- novonordiskfonden.dk
- Kind
- Foundation / funder
- Location
- Denmark
- Funding scale
- DKK 11.8bn (€1.6bn) awarded in 2025 across all areas (AI-specific totals not stated)
Flagship outputs
Focus areas
- ML for science
- Efficient ML
How to engage
- Grants / calls (open competition)
- Compute access via DCAI/Gefion
- Strategic partnership
Sources · verified November 2026
Anchor: #eu-novo-nordisk-foundation-ai
OpenEuroLLM consortium
Teams wanting open European-language training data, evaluation tooling and future open base models; organisations looking to join EU open-LLM work via partners or community channels.
High confidenceDetailsClose
EU Digital Europe consortium of about 20 universities, companies and EuroHPC centres (Feb 2025, 36 months) building open, EU-compliant multilingual foundation models for European languages; its main 9B 'Prelude' model is still marked work in progress on Hugging Face (Nov 2026).
Strengths
- Pan-European: about 20 partners spanning universities, research organisations, companies and EuroHPC centres
- Strategic EuroHPC compute allocation across four major supercomputers
- Open artefacts already public: MixtureVitae and MT-Nemotron-CC datasets, data catalogue, eval and training code
Watch-outs
- Main open model not yet released as of Nov 2026: flagship 9B 'Prelude' still marked work in progress
- Coordinator states securing enough compute for final models remains a 'significant challenge'
- Budget small vs industry labs; UK and Swiss entities excluded by EU funding rules; name overlap with EuroLLM
Leading researchers
- Project coordinator (Charles University, ÚFAL)Computational linguist; principal investigator for the coordinator
- Co-lead (AMD Silo AI)Co-founder of Silo AI; technical co-lead at launch
- Website
- openeurollm.eu
- Kind
- Network / society
- Institution
- Coordinated by Charles University (ÚFAL), Prague; co-led with AMD Silo AI
- Location
- Prague (coordinator); partners across Europe, Czechia (EU-wide consortium), Pan-European
- Founded
- 2025
- Funding scale
- €37.4m total, of which ~€20m from the EU Digital Europe Programme (grant 101195233); compute via EuroHPC strategic allocations on four supercomputers
Flagship outputs
Focus areas
- NLP / LLMs
- Efficient ML
How to engage
- Open-source models, data and code
- Community events (talks, winter school)
- Joint EU projects via consortium partners
Sources · verified November 2026
Anchor: #eu-openeurollm
Oxford Machine Learning (ELLIS Unit Oxford)
Companies and labs seeking vision, multi-agent RL, geometric deep learning or AI-safety collaborators and top-tier PhD/postdoc talent; engage via the specific department (CS, Engineering or Statistics) or ELLIS Unit Oxford.
Medium confidenceDetailsClose
Oxford's ML research spans Computer Science, Engineering (incl. the Visual Geometry Group) and Statistics; the ELLIS Unit Oxford (est. Sept 2020) links them with Warwick and the Ellison Institute of Technology.
Strengths
- Top-4 European ML-core output and #2 in AI venues (CSRankings, 2021–2025)
- World-class vision heritage (VGG; Zisserman, Vedaldi, Torr) plus RL/multi-agent (Foerster, Whiteson)
- Large ELLIS Unit: 15+ Fellows and dozens of members across Oxford and (since 2026) Warwick
- Strong AI-safety and trustworthy-AI research cluster
Watch-outs
- Low NLP venue output relative to peers (#55 in Europe in CSRankings NLP)
- Activity split across departments and colleges; partners must navigate multiple groups
- Several senior PIs hold parallel industry roles (e.g., Teh lists DeepMind on his ELLIS profile), which can limit availability
Leading researchers
- Director, ELLIS Unit Oxford (ELLIS Scholar)Statistical learning theory and optimisation
- Philip TorrTop 2%Professor, Engineering Science; ELLIS Fellowin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Oxford; AI & Image Processing; UK ICT #9). Vision and deep learning; top Oxford ML-core output in CSRankings 2021-25
- Michael BronsteinTop 2%Professor; ELLIS Program Director and Fellowin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Oxford; AI & Image Processing (Oxford rank 81)). Geometric deep learning and graph ML
- Yee Whye TehTop 2%Professor; ELLIS Program Director and Fellow (also DeepMind)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Oxford; AI & Image Processing (Oxford rank 77)). Bayesian nonparametrics and probabilistic ML; also at DeepMind
- Yarin GalTop 2%Faculty (OATML group); ELLIS Scholarin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Oxford; AI & Image Processing (Oxford rank 65)). Bayesian deep learning and AI safety
- Andrew ZissermanTop 2%Professor, Visual Geometry Groupin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Oxford; UK ICT #38). Computer vision (VGG networks); ranked #38 in UK ICT on the 2026 list
- Website
- ellis.ox.ac.uk
- Kind
- University group / lab
- Institution
- University of Oxford
- Location
- Oxford, United Kingdom
- Founded
- 2020
- CSRankings signal
- #4 in Europe for ML core (ICML/NeurIPS/ICLR), 2021–2025: 79.3 adjusted pubs; #2 for AI (AAAI/IJCAI), #8 for vision (CSRankings data)
Flagship outputs
Focus areas
- Computer vision
- RL
- Probabilistic ML
- Trustworthy / safe AI
- Theory
- ML for science
How to engage
- PhD / ELLIS PhD programme
- Industry partnership programme
- Spin-out / tech transfer
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-oxford-ellis-unit
Pioneer Centre for Artificial Intelligence (P1)
Companies and foundations seeking NLP, speech, causality or learning-theory partners in the Nordics, and PhD/postdoc hiring across five Danish universities; engage by collaboratory.
Medium confidenceDetailsClose
Denmark's national AI research centre (opened end-2021 with DKK 352m from five foundations), hosted by the University of Copenhagen with Aalborg, Aarhus, DTU and ITU, organised into collaboratories with 1,394 researchers listed.
Strengths
- National reach: co-led by five Danish universities with 1,394 researchers listed
- Europe-leading NLP output from Copenhagen (#4 in CSRankings 2021–2025)
- Foundation-backed: DKK 352.4m (€47m) from DNRF, Novo Nordisk, Carlsberg, Villum and Lundbeck foundations
- Thematic collaboratories give clear entry points by topic
Watch-outs
- Core-ML venue output (#18 in Europe) trails its NLP strength
- Multi-university structure can complicate contracting and IP ownership
- Foundation grant is time-limited (focal areas run 5–7 years from 2021); successor funding not stated in sources reviewed
Leading researchers
- Serge BelongieTop 2%Directorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Copenhagen; AI & Image Processing). Computer vision; leads the Pioneer Centre
- Isabelle AugensteinTop 2%Professor; Speech & Language collaboratory co-leadin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Copenhagen; AI & Image Processing). Fact-checking and explainable NLP
- Anders SøgaardTop 2%Professor, Univ. of Copenhagenin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Copenhagen; AI & Image Processing). NLP; top Copenhagen NLP output in CSRankings 2021-25
- Learning Theory & Optimisation collaboratory co-lead (DTU)Probabilistic ML; led Copenhagen ELLIS unit at 2020 launch
- Causality & eXplainability collaboratory co-leadGeometric and fair ML for medical imaging
- Website
- aicentre.dk
- Kind
- Research institute
- Institution
- University of Copenhagen (host) with DTU, ITU Copenhagen, Aalborg and Aarhus universities
- Location
- Copenhagen, Denmark
- Founded
- 2021
- CSRankings signal
- University of Copenhagen: #18 in Europe for ML core (21.9 adj.), #4 for NLP (ACL/EMNLP/NAACL), 2021–2025 (CSRankings data)
- Funding scale
- DKK 352.4m (€47m) from five foundations (DNRF, Novo Nordisk, Carlsberg, Villum, Lundbeck); focal areas run 5–7 years
Flagship outputs
Focus areas
- NLP / LLMs
- Causality
- Computer vision
- Theory
- Trustworthy / safe AI
How to engage
- PhD / ELLIS PhD programme
- Industry partnership programme
- Visiting researcher
- Foundation-funded calls
Sources · verified November 2026
Anchor: #eu-pioneer-centre-for-ai
Politecnico di Milano AI (DEIB AIRLab & RL3 Reinforcement Learning Lab)
Energy, Industry 4.0 and fintech companies wanting reinforcement-learning, online-learning or mechanism-design collaborators, and recruiters seeking RL PhDs in Italy.
Medium confidenceDetailsClose
PoliMi's ML strength is concentrated in DEIB's AI and Robotics Lab, notably the RL3 reinforcement-learning group, alongside online learning, algorithmic game theory and graph/learning-systems research. PoliMi is a partner in ELLIS Unit Milan (est. Oct 2021) with two ELLIS Fellows.
Strengths
- #14 in Europe for ML core from only 9 ML-active faculty (CSRankings 2021–2025), driven by RL
- ELLIS Unit Milan partner with 15 listed PoliMi members incl. Fellows Restelli and Alippi
- Applied RL track record with industry in green tech, Industry 4.0 and energy (lab site)
- Nine PoliMi authors in the AI & Image Processing subfield are on the 2026 Stanford–Elsevier list, incl. Cesa-Bianchi (#6 in Italy ICT) and Alippi
Watch-outs
- Narrow profile: weak in NLP (#94 in Europe) and vision (#37) in CSRankings
- Lab site names no specific industrial partners or spin-offs
- No institute-level AI front door or funding figures; engagement runs group by group
Leading researchers
- Faculty; ELLIS FellowReinforcement learning; top PoliMi ML-core output in CSRankings 2021-25
- Faculty; ELLIS Unit Milan co-directorAlgorithmic game theory and online learning
- Faculty; ELLIS ScholarRL theory; second-highest PoliMi ML-core output in CSRankings 2021-25
- Cesare AlippiTop 2%Faculty; ELLIS Fellow (also USI-SUPSI)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (Politecnico di Milano; AI & Image Processing; Italy ICT #21). Learning in non-stationary environments and graph-based learning
- Website
- rl.airlab.deib.polimi.it
- Kind
- University group / lab
- Institution
- Politecnico di Milano (Department of Electronics, Information and Bioengineering)
- Location
- Milan, Italy
- CSRankings signal
- #14 in Europe for ML core (ICML/NeurIPS/ICLR; 28.0 adj., 9 ML-active faculty), #9 for AI (AAAI/IJCAI), #19 combined, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- RL
- Theory
- Efficient ML
How to engage
- PhD / ELLIS PhD programme
- Industry research contracts
- Spin-out / tech transfer
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-polimi-airlab-rl
PR[AI]RIE-PSAI – Paris School of AI (IA Cluster)
Early-career researchers (AI Fellow posts), PhD candidates, and companies wanting industry partnerships or startup links with PSL, Paris Cité, Inria, CNRS and Institut Pasteur AI groups.
Medium confidenceDetailsClose
France 2030 'IA Cluster' (ANR: Sep 2024–Sep 2029, coordinator Isabelle Ryl) succeeding the PRAIRIE 3IA institute (2019); the largest IA Cluster award (€75m of €360m) for AI research chairs, fellows, PhDs and a bachelor-to-doctorate AI curriculum in Paris. The research groups at ENS and Collège de France are described separately (psl-ens-prairie-psai); Paris's other IA Cluster is Hi! PARIS (ip-paris-hi-paris).
Strengths
- Largest award in France's IA Cluster programme (€75m), with a 5-year horizon
- Combines strong Paris partners: PSL, Paris Cité, CNRS, Inria and Institut Pasteur (AI for life sciences)
- Concrete hiring routes: recurring AI Fellow calls and chair positions
- Explicit startup goal (50 deep-tech companies by 2030) with entrepreneurship programmes
Watch-outs
- Fragmented public information: leadership, industry partners and current numbers not consolidated on one site
- Fellow salaries (~€3.3–3.9k gross/month) are modest vs industry and some European peers
- It is a funding umbrella, not a single lab: the research itself sits in partner groups such as ENS/Collège de France (see psl-ens-prairie-psai), Inria and Institut Pasteur
Leading researchers
- Project coordinator, PR[AI]RIE-PSAI (ANR record)Coordinates the €75m IA Cluster award for Université PSL
- President, Université PSL (lead institution); elected interim president July 2024, then presidentSucceeded Alain Fuchs, who announced the 2024 award but resigned in June 2024
- Website
- prairie-institute.fr
- Kind
- Network / society
- Institution
- Université PSL (coordinator) with Université Paris Cité, CNRS, Inria and Institut Pasteur
- Location
- Paris, France
- Founded
- 2024
- Funding scale
- €75m France 2030 IA Cluster grant (ANR, Sep 2024–Sep 2029), the highest of the €360m programme
Flagship outputs
Focus areas
- Theory
- ML for science
- Computer vision
- NLP / LLMs
How to engage
- AI Fellow and chair recruitment
- PhD programme
- Industry partnership programme
- Spin-out / startup support (target 50 deep-tech startups by 2030)
Sources · verified November 2026
Anchor: #eu-prairie-psai
Sapienza University of Rome AI (Sapienza NLP, GLADIA & DIAG)
Partners needing Italian-language LLMs and multilingual semantics, model merging/reuse research, or computational social science on online information; public-sector and industrial NLP via Babelscape.
Medium confidenceDetailsClose
Sapienza's Computer Science and DIAG departments host Italy's leading academic NLP group (Navigli; Minerva Italian LLMs, BabelNet, spin-off Babelscape), the GLADIA representation-learning lab (Rodolà) and a computational social science group studying misinformation.
Strengths
- Built Italy's first from-scratch Italian LLM family (Minerva, up to 7B) with national HPC and an industrial partner
- Balanced AI output: #10 AI, #15 NLP and #34 ML core in Europe (CSRankings 2021–2025)
- Proven spin-out path: Babelscape commercialises Sapienza multilingual NLP and Minerva
- Active representation-learning lab (GLADIA) with regular ICLR/ICML/NeurIPS output
Watch-outs
- No ELLIS unit in Rome (Italian units are Milan, Genoa, Modena, Trento, Turin)
- Minerva and Sapienza NLP depend on time-limited PNRR/FAIR recovery-plan funding
- AI groups spread across departments with no single institute website, headcount or budget
Leading researchers
- Full Professor, DIAG (Dept. of Computer, Control and Management Engineering); leads Sapienza NLP and the Minerva LLM project; co-founder of BabelscapeCreator of BabelNet; ERC Starting (MultiJEDI) and Consolidator (MOUSSE) grants; top Sapienza NLP output in CSRankings
- Full Professor; PI of the GLADIA lab (23 current members)Model merging, latent-space communication, geometric deep learning; ERC Starting Grant SPECGEO (2018)
- Walter QuattrociocchiTop 2%Full Professor, Dept. of Computer Science; head of the Center of Data Science and Complexity for Society (CDCS)in the 2026 list per the topcitedscientists.com mirror (Sapienza; AI & Image Processing, subfield #3,003). Misinformation and online polarisation
- Website
- nlp.uniroma1.it
- Kind
- University group / lab
- Institution
- Sapienza University of Rome
- Location
- Rome, Italy
- CSRankings signal
- #34 in Europe for ML core (10.4 adj., 13 faculty), #15 for NLP, #10 for AI (AAAI/IJCAI), 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- NLP / LLMs
- Efficient ML
- Theory
- Trustworthy / safe AI
How to engage
- Industry deployment via spin-off Babelscape
- National research partnerships (FAIR programme, CINECA compute)
- PhD programme
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-sapienza-rome-ai
Swiss AI Initiative & Swiss National AI Institute (SNAI)
Teams wanting a fully open (Apache 2.0) multilingual LLM with published data and recipes, Swiss researchers seeking large GPU allocations, and companies seeking an industry partnership with ETH/EPFL.
High confidenceDetailsClose
National open-science programme (launched Dec 2023) building fully open foundation models on CSCS's Alps supercomputer; run by SNAI, the ETH–EPFL joint institute set up in Oct 2024. Builder of the Apertus open LLMs.
Strengths
- Among the most open LLM efforts anywhere: weights, training data and recipes public under Apache 2.0 (Apertus)
- Dedicated national compute: 10m+ GPU hours on CSCS Alps, a 10,000+ GH200 GPU system
- Broad base: 800+ researchers incl. ~70 AI professors from 10+ Swiss institutions
- Strong multilingual focus incl. Swiss German and Romansh; Apertus 1.5 adds image and audio understanding
Watch-outs
- Explicitly not aiming at frontier performance: co-lead says Apertus is 'not about competing' with private frontier models
- Calls and compute are mainly for Swiss-based researchers; Switzerland is excluded from some EU-funded consortia
- Programme structure (many projects across institutions) makes single points of contact harder to identify
Leading researchers
- Executive Co-Director, SNAI (Managing Director, ETH AI Center), per the Oct 2024 founding announcementCo-leads the joint ETH–EPFL institute
- Executive Co-Director, SNAI (Managing Director, EPFL AI Center), per the Oct 2024 founding announcementCo-leads the joint ETH–EPFL institute
- Professor, EPFL; Apertus project co-leadOptimisation and distributed/efficient training of LLMs
- Professor, EPFL; Apertus project co-leadNLP, knowledge and reasoning in language models
- Research scientist, ETH Zurich; Apertus project co-leadLeads Apertus pretraining work at ETH
- Website
- swiss-ai.org
- Kind
- Network / society
- Institution
- ETH Zurich and EPFL (ETH AI Center + EPFL AI Center), with CSCS
- Location
- Zurich / Lausanne / Lugano, Switzerland
- Founded
- 2023
- Funding scale
- CHF 20m from the ETH Board for 2025–2028, plus ETH/EPFL contributions and third-party funds; seeded with 10m+ GPU hours on Alps (10,000+ GH200 GPUs)
Flagship outputs
Focus areas
- NLP / LLMs
- Efficient ML
- Trustworthy / safe AI
- ML for science
How to engage
- Grants / calls (recurring Large Project Calls; the 4th ran 3 Aug–14 Sep 2026)
- Compute calls (Alps GPU hours, for Swiss-based researchers)
- Industry partnership programme (Swisscom is strategic partner)
- Open-source models and code
Sources · verified November 2026
Anchor: #eu-swiss-ai-initiative-snai
The Alan Turing Institute
Public-sector, defence and security partners, and UK university networks; others should check current strategy before committing.
Medium confidenceDetailsClose
UK national institute for data science and AI (charity, founded 2015 by five universities and EPSRC). Since 2025 it has been restructuring toward defence and national security under government direction.
Strengths
- National convening role with 13 member universities and an open university network (2023)
- Long-standing defence/security partnerships (GCHQ, MoD) now central to its mandate
- Government funding line (£100m over five years from 2024)
- Hosted at the British Library in central London
Watch-outs
- Governance turbulence: staff no-confidence letter (Dec 2024), CEO exit (Sep 2025), interim CEO into 2026, and chair Doug Gurr resigned (Apr 2026; Vanessa Lawrence interim chair)
- April 2026 UKRI review found underperformance on strategy and value for money; long-term funding under review
- Redundancy rounds (~140 of 440 roles consulted in 2024) reduce research continuity
Leading researchers
- Interim CEOStatistical ML; led the Institute after the 2025 CEO departure
- CEO-designate (announced Feb 2026; expected to start May 2026, start not confirmed in sources reviewed)Former head of the UK Government Communications Centre
- Website
- turing.ac.uk
- Kind
- Research institute
- Location
- London, United Kingdom
- Founded
- 2015
- Funding scale
- £100m over five years from the 2024 Spring Budget (health, environment, defence); initial £42m + £5m per founder university
Focus areas
- Trustworthy / safe AI
- ML for science
- Probabilistic ML
How to engage
- Strategic partnership
- Grants / calls
- University network membership
- Fellowships
Sources · verified November 2026
Anchor: #eu-alan-turing-institute
TU Darmstadt AI & hessian.AI
Companies needing NLP/LLM, robot-learning or vision collaborators, and SMEs or start-ups wanting compute, consulting and training via hessian.AI's AI Service Center.
Medium confidenceDetailsClose
TU Darmstadt's NLP, ML, vision and robot-learning groups anchor hessian.AI (est. Aug 2020, co-directed by Mira Mezini and Kristian Kersting), a state-backed centre spanning 13 Hessian universities with an AI Service Center and the 'Reasonable AI' Cluster of Excellence.
Strengths
- Top-6 European NLP output (UKP Lab) in CSRankings 2021–2025
- Robot learning heritage (Peters) combined with vision (Roth) and neuro-symbolic ML (Kersting)
- hessian.AI provides SME-facing compute, consulting and a start-up programme
- Cluster of Excellence 'Reasonable AI' on resource-efficient, privacy-aware learning
Watch-outs
- ML-core output (#20 in Europe) is modest relative to NLP strength
- State-network structure (13 universities) can blur who to contract with
- Core state funding was project-based (€38m build-up phase to 2024, then a €12m notice); long-term permanence not stated
Leading researchers
- Iryna GurevychTop 2%Professor, UKP Labin the V9 (Aug 2026) list per the topcitedscientists.com mirror (TU Darmstadt; AI & Image Processing). NLP; one of Europe's highest NLP-venue outputs in CSRankings 2021-25
- Kristian KerstingTop 2%Professor, AI & ML Lab; Co-director, hessian.AIin the V9 (Aug 2026) list per the topcitedscientists.com mirror (TU Darmstadt; AI & Image Processing). Probabilistic/neuro-symbolic ML; top Darmstadt ML-core output
- Jan PetersTop 2%Professor, Intelligent Autonomous Systemsin the V9 (Aug 2026) list per the topcitedscientists.com mirror (TU Darmstadt; subfield Industrial Engineering & Automation). Robot learning and RL
- Professor, Visual InferenceComputer vision; led the Darmstadt ELLIS Unit at its 2020 launch
- Website
- hessian.ai
- Kind
- University group / lab
- Institution
- TU Darmstadt (hessian.AI spans 13 Hessian universities)
- Location
- Darmstadt, Germany
- Founded
- 2020
- CSRankings signal
- TU Darmstadt: #20 in Europe for ML core (20.1 adj.), #6 for NLP, #14 for vision; #15 combined, 2021–2025 (CSRankings data)
- Funding scale
- Hesse state: €38m for the 2020–2024 build-up (20 professorships) plus a €12m funding notice (Dec 2024); federal funding for the AI Service Center
Flagship outputs
Focus areas
- NLP / LLMs
- Robotics learning
- Computer vision
- Efficient ML
- Trustworthy / safe AI
How to engage
- PhD / ELLIS PhD programme
- AI Service Center (compute, consulting, training)
- Spin-out support (AI Startup Rising)
- Industry partnership programme
Sources · verified November 2026
Anchor: #eu-tu-darmstadt-hessian-ai
TU Delft AI Initiative (TU Delft AI Labs & ELLIS Unit Delft)
Engineering-heavy companies (mobility, energy, infrastructure, finance) wanting applied ML, sequential decision-making, multi-agent or safe-AI collaborators, or co-funded ICAI labs; recruiters targeting the AI Labs PhD cohort.
Medium confidenceDetailsClose
TU Delft's university-wide AI programme of 24 'AI + domain' labs (50+ lab directors, 96 PhD candidates) and ELLIS Unit Delft, one of ELLIS's 17 founding units (2019) and now its fourth-largest, with 40+ researchers across four faculties.
Strengths
- Scale of structured AI hiring: 24 AI Labs with 50+ lab directors and 96 PhD candidates
- ELLIS Unit Delft: founding unit (2019), fourth-largest in the network, 7 Fellows and 5 Scholars; co-directs 11 of the 24 AI Labs
- #6 in Europe for AAAI/IJCAI output and #19 for ML core (CSRankings 2021–2025)
- Established industry-lab model via ICAI (e.g., labs with ING, DSM and ProRail)
Watch-outs
- Output is spread thin: 29 ML-active faculty but no individual with more than ~2 adjusted ML-core papers in 2021–2025 (CSRankings)
- Labs pair AI with engineering domains, so fundamental ML depth is thinner than at top ML centres
- AI Labs launch year and programme funding not stated on the pages reviewed; former ELLIS co-director Jens Kober now listed at Stuttgart
Leading researchers
- Director, ELLIS Unit Delft (ELLIS Fellow)Multi-agent decision making and RL
- Co-director, ELLIS Unit Delft (ELLIS Scholar)Autonomous-driving perception and datasets
- Matthijs SpaanTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (TU Delft; AI & Image Processing). Planning under uncertainty and safe RL; top TU Delft ML-core output in CSRankings 2021-25
- Faculty, Computer Vision LabDeep learning for vision; top TU Delft combined CSRankings output
- Faculty; lead of the ELIAS virtual centre (ELLIS Delft)Algorithmic game theory and optimisation for energy systems
- Website
- tudelft.nl
- Kind
- University group / lab
- Institution
- Delft University of Technology
- Location
- Delft, Netherlands
- Founded
- 2019
- CSRankings signal
- #19 in Europe for ML core (ICML/NeurIPS/ICLR; 20.3 adj., 29 ML-active faculty), #6 for AI (AAAI/IJCAI), #23 combined, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- RL
- Trustworthy / safe AI
- Probabilistic ML
- Computer vision
- Robotics learning
How to engage
- PhD / ELLIS PhD programme
- ICAI co-funded industry lab
- Industry partnership programme
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-tu-delft-ai
TU Wien Center for Artificial Intelligence and Machine Learning (CAIML)
Companies needing knowledge representation and reasoning, constraint solving and combinatorial optimisation, neuro-symbolic or explainable AI (e.g., scheduling, configuration, verification), and PhD hires via the iCAIML doctoral college.
Medium confidenceDetailsClose
TU Wien's cross-department AI centre, founded in late 2021 to bundle symbolic AI, machine learning and explainable AI / digital humanism; a 12-member board from Informatics and Mathematics runs it, with the iCAIML doctoral college. TU Wien has Europe's highest AAAI/IJCAI output in CSRankings 2021–2025.
Strengths
- #1 in Europe for AAAI/IJCAI output in CSRankings 2021–2025 (32.3 adjusted papers)
- Rare depth in symbolic AI (KR, logic, constraint solving) combined with ML and explainable AI under one centre
- Eight TU Wien authors in the AI & Image Processing subfield appear in the 2026 Stanford–Elsevier list, incl. Lukasiewicz and Eiter
- Structured doctoral training via the iCAIML doctoral college
Watch-outs
- Modest ML-core output (#21 in Europe) and little computer vision (#46)
- Young (founded 2021) coordinating centre; its 12-member board is not named on its website and no funding figures are published
- Not an ELLIS Unit site in sources reviewed; flagship outputs are papers and tools rather than widely adopted models
Leading researchers
- Thomas LukasiewiczTop 2%Faculty, TU Wien Informaticsin the V9 (Aug 2026) list per the topcitedscientists.com mirror (TU Wien; AI & Image Processing). Neuro-symbolic AI and reasoning with LLMs; top TU Wien ML-core output in CSRankings 2021-25
- Thomas EiterTop 2%Faculty, Knowledge-Based Systemsin the V9 (Aug 2026) list per the topcitedscientists.com mirror (TU Wien; AI & Image Processing). Knowledge representation and answer-set programming
- Faculty, Institute of Logic and Computation; presented CAIML at its 2021 openingArgumentation and computational logic
- Faculty, Algorithms and ComplexitySAT/constraint solving and parameterised complexity; highest TU Wien AAAI/IJCAI output in CSRankings 2021-25
- Faculty, Institute of Analysis and Scientific Computing; presented CAIML at its 2021 openingReinforcement learning and scientific ML
- Website
- caiml.org
- Kind
- University group / lab
- Institution
- TU Wien (Vienna University of Technology)
- Location
- Vienna, Austria
- Founded
- 2021
- CSRankings signal
- #21 in Europe for ML core (ICML/NeurIPS/ICLR; 19.5 adj., 21 ML-active faculty) but #1 for AI (AAAI/IJCAI; 32.3 adj.); #16 combined, 2021–2025 (CSRankings data)
Focus areas
- Theory
- Trustworthy / safe AI
- NLP / LLMs
How to engage
- PhD (iCAIML doctoral college)
- Industry partnership programme
- Visiting researcher
- Summer schools
Sources · verified November 2026
Anchor: #eu-tu-wien-caiml
Tübingen AI Center & Cluster of Excellence 'Machine Learning for Science'
Science-driven companies and funders seeking ML-for-science, probabilistic-numerics or robust-ML collaboration, PhD talent via IMPRS-IS/ELLIS, and partners for open multilingual LLM evaluation (OpenEuroLLM).
High confidenceDetailsClose
Federally designated AI competence centre (2018) hosted by the University of Tübingen, with 25 ML groups and 300+ PhD students and postdocs. Its DFG Cluster of Excellence 'Machine Learning for Science' (since 2019) was renewed for seven years from 2026.
Strengths
- #7 European ML-core output in CSRankings 2021–2025 with 19 active ML faculty
- Cluster of Excellence renewed for 2026–2032 with MPI-IS, MPI Biological Cybernetics, ELLIS Institute Tübingen, IWM and AIMS as partners
- Distinctive ML-for-science agenda and probabilistic numerics expertise
- Role in OpenEuroLLM: training and evaluating highly multilingual open foundation models
Watch-outs
- Small-city location; heavily intertwined with MPI-IS and ELLIS Institute, so affiliations can be confusing
- Low presence in classic AI venues (AAAI/IJCAI) and modest NLP output
- Public-university contracting and IP terms can be slow for start-ups
Leading researchers
- Ulrike von LuxburgTop 2%Professor; Cluster of Excellence co-spokesperson (with Philipp Berens)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under Tübingen AI Center; Germany ICT #15). Statistical learning theory and explainability
- Philipp HennigTop 2%Professor, Methods of MLin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Tübingen; AI & Image Processing). Probabilistic numerics
- Matthias HeinTop 2%Professor, Machine Learningin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Tübingen; AI & Image Processing). Adversarial robustness and reliable deep learning
- Andreas GeigerTop 2%Professorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Tübingen; AI & Image Processing (Tübingen rank 2)). Autonomous driving and 3D vision
- Matthias BethgeTop 2%Professorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Tübingen; AI & Image Processing). Neuro-inspired and robust vision; open model evaluation
- Website
- tuebingen.ai
- Kind
- University group / lab
- Institution
- University of Tübingen (with MPI for Intelligent Systems)
- Location
- Tübingen, Germany
- Founded
- 2018
- CSRankings signal
- University of Tübingen: #7 in Europe for ML core (ICML/NeurIPS/ICLR), 50.0 adj. pubs, 19 faculty, 2021–2025; #9 for vision (CSRankings data)
- Funding scale
- DFG Excellence Strategy cluster funded for 7 years from 2026 (amount not stated)
Flagship outputs
Focus areas
- ML for science
- Probabilistic ML
- Trustworthy / safe AI
- Computer vision
- NLP / LLMs
- Theory
How to engage
- PhD / ELLIS PhD programme
- PhD (IMPRS-IS)
- Cyber Valley industry partnership
- Spin-out / tech transfer
Sources · verified November 2026
Anchor: #eu-tuebingen-ai-center
UCL Centre for Artificial Intelligence & Gatsby Computational Neuroscience Unit
Recruiters and research partners seeking PhD/postdoc talent with deep theoretical ML training, kernel/probabilistic-ML collaborators, and UKRI CDT industry partnerships.
Medium confidenceDetailsClose
UCL's foundational AI centre (vision, ML, NLP, machine action) runs the UKRI CDT in Foundational AI; the Gatsby Unit, funded by the Gatsby Charitable Foundation, joins ML theory and neuroscience. UCL hosts ELLIS Unit London (est. Sept 2020; directors Gretton, Guedj, Briol).
Strengths
- Exceptional talent pipeline: Gatsby alumni co-founded Google DeepMind; ~30% of alumni continue research in companies such as Google DeepMind, Meta and Anthropic
- Hosts ELLIS Unit London; Gatsby joined it in 2020
- Theory-heavy ML (learning theory, probabilistic ML) combined with computational neuroscience
- Close links with the Alan Turing Institute and Sainsbury Wellcome Centre
Watch-outs
- CSRankings undercounts Gatsby (neuroscience-oriented faculty), so its ML-core rank (#13) may understate impact
- Gatsby's PhD intake is small and highly selective
- Low NLP venue output relative to peers (#52 in Europe)
Leading researchers
- Massimiliano PontilTop 2%Professor, UCL; also Istituto Italiano di Tecnologia (IIT)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (affiliation listed as Istituto Italiano di Tecnologia; Italy ICT #69). Statistical learning theory, multi-task learning
- Jun WangTop 2%Professorin the V9 (Aug 2026) list per the topcitedscientists.com mirror (listed under UCL Centre for AI; UK ICT #99). RL, multi-agent and LLM agents; top UCL ML-core output in CSRankings 2021-25
- ProfessorData-efficient ML, Gaussian processes; 'Mathematics for ML' co-author
- ProfessorProbabilistic ML
- Professor3D vision; co-founder of Synthesia
- Arthur GrettonTop 2%Co-director, ELLIS Unit London (Gatsby Unit)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (UCL; AI & Image Processing). Kernel methods and statistical testing
- Website
- ucl.ac.uk
- Kind
- University group / lab
- Institution
- University College London
- Location
- London, United Kingdom
- CSRankings signal
- #13 in Europe for ML core (28.2 adj.), #12 for vision, 2021–2025 (CSRankings data; Gatsby faculty largely not CSRankings-listed)
Flagship outputs
Focus areas
- Probabilistic ML
- RL
- Theory
- Computer vision
- NLP / LLMs
How to engage
- PhD / ELLIS PhD programme
- UKRI CDT industry partnership
- Spin-out / tech transfer (UCL Business)
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-ucl-ai-centre-gatsby
Universitat Pompeu Fabra AI (AI & ML group, Music Technology Group, COLT)
Partners in bandits/online and statistical learning theory, RL and planning, emergent communication and computational linguistics, and audio/music AI; ICREA chairs make it attractive for senior hires.
Medium confidenceDetailsClose
UPF combines learning theory and online learning (Lugosi, Neu), reinforcement learning and planning (AI & ML group), computational linguistics (COLT) and the Music Technology Group (founded 1994), and is a partner in ELLIS Unit Barcelona (est. March 2023).
Strengths
- Internationally known learning-theory group (Lugosi, ELLIS Fellow; Neu, ELLIS Scholar)
- Six UPF authors in the 2026 Stanford–Elsevier Spain ICT top-100 view, more than any other Barcelona institution
- Long-running Music Technology Group (since 1994) with widely used open-source audio tooling (Essentia)
- Member of ELLIS Unit Barcelona alongside CVC, BSC, UPC, IIIA-CSIC and others
Watch-outs
- Not covered by CSRankings, so ML venue output cannot be benchmarked
- AI expertise spread across engineering, economics and linguistics departments with no single institute
- Small core-ML group (AI & ML group lists 6 faculty plus postdocs); limited vision or large-model capacity
Leading researchers
- Gábor LugosiTop 2%ICREA Research Professor, Dept. of Economics and Business, UPF (also Barcelona School of Economics); ELLIS Fellowlisted on the topcitedscientists.com Spain ICT page (#68, shown under Barcelona School of Economics). Statistical learning theory, prediction and learning in games
- Marco BaroniTop 2%ICREA Research Professor, Department of Linguistics, UPFlisted on the topcitedscientists.com Spain ICT page (UPF; top UPF author). Leads ERC project ALiEN on emergent communication in deep networks
- Xavier SerraTop 2%Full Professor, Dept. of Engineering, UPF; founder (1994) and director of the Music Technology Group; director of the UPF-BMAT Chair on AI and Musiclisted on the topcitedscientists.com Spain ICT page (UPF). Audio signal processing and music information retrieval
- Full Professor; leader of the UPF AI & ML research group; ELLIS memberReinforcement learning and automated planning
- Website
- upf.edu
- Kind
- University group / lab
- Institution
- Universitat Pompeu Fabra (UPF)
- Location
- Barcelona, Spain
- CSRankings signal
- Not ranked: UPF is not listed as an institution in CSRankings, so its ML output is uncounted (CSRankings data, 2026-11)
Flagship outputs
Focus areas
- Theory
- RL
- NLP / LLMs
- ML for science
How to engage
- PhD / ELLIS PhD programme
- ICREA research professorships (senior recruitment)
- Industry collaboration / licensing (e.g. MTG audio technologies)
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-upf-barcelona-ai
University of Bonn & Lamarr Institute for Machine Learning and AI (with TU Dortmund)
Industrial and logistics firms seeking applied, resource-aware or trustworthy ML partners, robotics/SLAM and agricultural-robotics collaborators, and Fraunhofer-backed transfer projects in North Rhine-Westphalia.
Medium confidenceDetailsClose
The Lamarr Institute grew out of the ML2R competence centre in 2022 as one of Germany's five university AI centres, joining the University of Bonn, TU Dortmund and two Fraunhofer institutes; Bonn is especially strong in robot perception and embodied AI.
Strengths
- Long-term federal + state funding (~€126m initially to 2028) for one of Germany's five university AI centres
- Europe-leading robot perception and embodied AI in Bonn (#5 in Europe for robotics venues, CSRankings 2021–2025)
- Built-in transfer route through two Fraunhofer institutes (IAIS, IML) for industry and logistics
- Bonn is part of the new ELLIS Unit NRW (est. 1 Apr 2026, 16 institutions): Gall is one of its eight directors and Behnke a Fellow
Watch-outs
- Modest ML-core venue output: Bonn #86 and TU Dortmund #63 in Europe (CSRankings 2021–2025)
- Multi-site, four-partner structure (two universities, two Fraunhofer institutes) can complicate contracting
- Institute website does not state headcount, founding details or post-2028 funding
Leading researchers
- Cyrill StachnissTop 2%Professor, Photogrammetry & Robotics Lab, University of Bonn; PhenoRob cluster spokesperson; Lamarr-affiliatedlisted on the topcitedscientists.com Universität Bonn page (Industrial Engineering & Automation). SLAM and robot perception
- Chair of Autonomous Intelligent Systems, University of Bonn; Lamarr area chair Embodied AI; ELLIS FellowNimbRo team won RoboCup, MBZIRC and the ANA Avatar XPRIZE
- Professor, Computer Vision Group, University of Bonn; one of eight directors of ELLIS Unit NRW (est. 1 Apr 2026)Action recognition, video understanding and human pose; co-author of SemanticKITTI
- Director, Lamarr Institute (Fraunhofer IAIS / University of Bonn)Leads the institute's hybrid 'Triangular AI' agenda combining data with knowledge and context
- Website
- lamarr-institute.org
- Kind
- University group / lab
- Institution
- University of Bonn; TU Dortmund University; Fraunhofer IAIS and IML
- Location
- Bonn / Dortmund / Sankt Augustin, Germany
- Founded
- 2022
- CSRankings signal
- University of Bonn: #5 in Europe for robotics (ICRA/IROS/RSS, 16.0 adj.), #29 vision, #86 ML core; TU Dortmund #63 ML core, #15 AI, 2021–2025 (CSRankings data)
- Funding scale
- About €126m from the federal government and North Rhine-Westphalia, initially through 2028
Focus areas
- Robotics learning
- Computer vision
- Efficient ML
- Trustworthy / safe AI
- NLP / LLMs
How to engage
- PhD / ELLIS PhD programme
- Industry transfer projects (via Fraunhofer IAIS and IML)
- Lamarr Fellows and conference/events
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-bonn-lamarr-institute
University of Edinburgh School of Informatics (ELLIS Unit Edinburgh)
Companies needing NLP/LLM, speech or robot-learning collaborators, open multilingual LLM partners (EuroLLM), or a large PhD pipeline via UKRI AI CDTs.
High confidenceDetailsClose
One of Europe's largest informatics schools (formed 1998 from the former Department of AI and other units; 370 academic staff). Its ELLIS Unit (with Heriot-Watt, est. 2020) spans NLP, ML foundations, RL and robot learning.
Strengths
- #1 in Europe for NLP venue output and #4 for AI venues (CSRankings, 2021–2025)
- Broadest ML bench in Europe: 41 CSRankings-listed faculty with ML-core papers
- Leads or co-leads three UKRI AI Centres for Doctoral Training (announced 2023)
- Contributor to EuroLLM, an open multilingual European LLM family
Watch-outs
- ML-core output per faculty member is lower than Oxbridge/ETH peers
- Large school (370 academic staff, six research institutes); partners need a specific institute or PI as the entry point
- UK post-Brexit status adds friction for some EU programme roles
Leading researchers
- Director, ELLIS Unit EdinburghNLP and interpretable/structured language models
- Professor; ELLIS FellowNLP and summarisation; high NLP output in CSRankings
- Timothy HospedalesTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Edinburgh; AI & Image Processing). Meta-learning and transfer; top Edinburgh ML-core output in CSRankings 2021-25
- Professor; ELLIS FellowRobot learning and embodied AI
- Chris WilliamsTop 2%Professor; ELLIS Fellowin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Edinburgh; AI & Image Processing). Gaussian processes and probabilistic ML
- Website
- informatics.ed.ac.uk
- Kind
- University group / lab
- Institution
- University of Edinburgh
- Location
- Edinburgh, United Kingdom
- Founded
- 1998
- CSRankings signal
- #8 in Europe for ML core (47.3 adj.), #1 for NLP (ACL/EMNLP/NAACL), #4 for AI; most ML-active faculty (41) in Europe, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- NLP / LLMs
- Probabilistic ML
- Robotics learning
- RL
- Computer vision
- Trustworthy / safe AI
How to engage
- PhD / ELLIS PhD programme
- UKRI Centres for Doctoral Training
- Industry partnership programme
- Spin-out / tech transfer
Sources · verified November 2026
Anchor: #eu-edinburgh-informatics
University of Freiburg Machine Learning Lab (AutoML) & ELLIS Unit Freiburg
Firms needing AutoML, hyperparameter optimisation or tabular foundation models, and investors tracking structured-data AI spin-outs; note Hutter's time is split with Prior Labs/SAP and ELLIS Tübingen.
High confidenceDetailsClose
Global centre for AutoML (Auto-sklearn, TabPFN) led by Frank Hutter, whose TabPFN spin-out Prior Labs SAP agreed to acquire in 2026; the lab now also lists an interim professor. The ELLIS Unit Freiburg (est. Dec 2019) adds vision and robotics.
Strengths
- World-leading AutoML group: Auto-sklearn and TabPFN are widely used open-source tools
- Exceptional spin-out signal: Prior Labs (TabPFN) acquisition by SAP with >€1bn investment plan
- Hutter co-initiated the AutoML conference and holds an ERC Consolidator Grant
- Complementary vision (Brox) and robot-learning (Valada, Boedecker) groups
Watch-outs
- Heavy dependence on one PI, who now holds multiple affiliations (lab lists an interim professor)
- Small number of ML-core faculty (4) in CSRankings
- Narrow focus (AutoML/tabular) relative to broad AI centres
Leading researchers
- Frank HutterTop 2%Professor; head of ELLIS Unit Freiburg; also ELLIS Institute Tübingen PI and Prior Labs CEOin the V9 (Aug 2026) list per the topcitedscientists.com mirror (affiliation listed as Prior Labs; Germany ICT #14). AutoML; TabPFN (Nature 2025); ERC CoG; also ELLIS Institute Tübingen
- Thomas BroxTop 2%Professor; ELLIS Fellowin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Freiburg; AI & Image Processing). Computer vision (U-Net, FlowNet lineage)
- Abhinav ValadaTop 2%Professor; ELLIS Scholarin the V9 (Aug 2026) list per the topcitedscientists.com mirror (Univ. Freiburg; subfield Industrial Engineering & Automation). Robot learning and perception
- ProfessorNeurorobotics and reinforcement learning
- Kind
- University group / lab
- Institution
- University of Freiburg
- Location
- Freiburg, Germany
- Founded
- 2019
- CSRankings signal
- #27 in Europe for ML core (13.0 adj.) from 4 ML-active faculty, 2021–2025 (CSRankings data)
- Funding scale
- Spin-out Prior Labs: SAP to acquire and invest >€1bn over four years (announced May 2026; reported completed July 2026)
Flagship outputs
Focus areas
- Efficient ML
- Robotics learning
- Computer vision
- RL
How to engage
- PhD / ELLIS PhD programme
- Spin-out / tech transfer
- Industry collaboration / licensing (open-source AutoML tools)
Sources · verified November 2026
Anchor: #eu-freiburg-ml-lab
University of Manchester AI (ELLIS Unit Manchester, Centre for AI Fundamentals & NaCTeM)
Health, life-science and pharma partners needing probabilistic ML, Bayesian experimental design or biomedical text mining; NHS-linked translational projects; PhD hiring through a UKRI AI CDT run with Cambridge.
Medium confidenceDetailsClose
Manchester's ELLIS unit (established Sept 2021, announced Nov 2021) focuses on probabilistic modelling, human-AI collaboration and ML for health, twinned with Helsinki via Samuel Kaski's shared chair; the university also hosts the Centre for AI Fundamentals and NaCTeM, the UK's National Centre for Text Mining.
Strengths
- Clear health and life-science focus: ELLIS unit tied to the Pankhurst Institute and NHS partners
- Twinned with ELLIS Finland through Kaski's shared chair and Turing AI World-Leading Researcher Fellowship
- Europe top-15 NLP output (#14, CSRankings 2021–2025) anchored by NaCTeM's long-running biomedical text mining
- Strong PhD pipeline via a UKRI AI CDT co-run with Cambridge
Watch-outs
- Mid-table ML-core output (#43 in Europe, CSRankings 2021–2025)
- Centre for AI Fundamentals site gives no founding date, director or headcount; funding totals not published
- NaCTeM's public open-source footprint on GitHub is small; key leaders hold split appointments (Aalto, KU Leuven)
- Leadership is reported inconsistently: ELLIS lists Kaski and Rattray as unit directors, while the unit's own site shows Mauricio Álvarez López under 'Director'
Leading researchers
- Professor of AI, University of Manchester, and Research Director, Christabel Pankhurst Institute; Director of FCAI and ELLIS Unit Helsinki (Aalto); ELLIS Unit Manchester co-directorTuring AI World-Leading Researcher Fellowship on human-AI teams for drug design and digital twins
- Professor of Computational and Systems Biology, University of Manchester; ELLIS Unit Manchester co-director; directed the Institute for Data Science and AI at the unit's 2021 launchProbabilistic ML for genomics and health; ELLIS Health programme Fellow
- Professor, Chair in NLP and Text Mining; Director of NaCTeM; ELLIS Fellow (2024)Biomedical text mining; over 400 peer-reviewed papers; tools TerMine, AcroMine, RobotAnalyst
- Alejandro F. FrangiTop 2%Professor, University of Manchester (also KU Leuven); ELLIS Fellowin the 2026 list per the topcitedscientists.com mirror (Manchester; Nuclear Medicine & Medical Imaging, subfield #166). Computational medicine
- Website
- ellismcr.org
- Kind
- University group / lab
- Institution
- The University of Manchester
- Location
- Manchester, United Kingdom
- Founded
- 2021
- CSRankings signal
- #43 in Europe for ML core (8.2 adj., 12 faculty), #14 for NLP, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Probabilistic ML
- ML for science
- NLP / LLMs
- Trustworthy / safe AI
How to engage
- PhD (UKRI AI CDT in Decision Making for Complex Systems, with Cambridge) / ELLIS PhD programme
- Health partnerships (Christabel Pankhurst Institute, Christie NHS Foundation Trust)
- Text-mining services (NaCTeM)
- Summer schools and seminars
Sources · verified November 2026
Anchor: #eu-manchester-ellis-ai-fun
University of Sheffield NLP Group (with the GATE team and Centre for Machine Intelligence)
Media, legal, public-sector and verification organisations needing NLP for disinformation, legal text or social media; teams wanting an established open-source text-analytics platform (GATE) and EU-project partners.
Medium confidenceDetailsClose
Established in 1993, Sheffield's NLP group (23 listed staff) is one of the UK's largest language-processing groups and the home of the open-source GATE text-engineering platform; its recent work spans efficient and multilingual LLMs, legal NLP and disinformation detection.
Strengths
- Top-5 European NLP output (#5, CSRankings 2021–2025), ahead of most larger universities
- Three-decade NLP group (est. 1993) with 14 academic staff and the long-maintained GATE platform
- Applied strength in disinformation, social media and legal NLP with repeated EU-project leadership
- University Centre for Machine Intelligence adds an AI research-engineering team and spin-out support
Watch-outs
- Narrow focus: little ML-core (#80 in Europe), vision or robotics output
- Not part of an ELLIS unit; the Sheffield ELLIS scholar listed sits in Manchester's unit
- No group members found in the visible Stanford–Elsevier top-100 for Sheffield; group funding not published
Leading researchers
- Professor of Natural Language Processing and Head of the NLP research group, School of Computer ScienceTop Sheffield NLP output in CSRankings 2021–2025; LM adaptation, efficiency (pruning/quantisation), legal NLP
- Senior Lecturer and School Programmes Lead (UG); NLP group member (the group page still lists him as head, but newer profiles show Aletras leads the group)Information access, text reuse and biomedical text processing
- Senior Lecturer (Associate Professor) in NLP; GATE team memberDisinformation detection and LLM safety; Sheffield PI on Horizon Europe VIGILANT
- Website
- sheffield.ac.uk
- Kind
- University group / lab
- Institution
- The University of Sheffield
- Location
- Sheffield, United Kingdom
- Founded
- 1993
- CSRankings signal
- #5 in Europe for NLP (ACL/EMNLP/NAACL, 24.4 adj.), #80 for ML core (3.8 adj.), 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- NLP / LLMs
- Trustworthy / safe AI
- Efficient ML
How to engage
- EU Horizon consortium partnering (e.g. VIGILANT, EDMO)
- Industry collaboration / licensing around GATE
- PhD programme
- Spin-out and AI research engineering support (Centre for Machine Intelligence)
Sources · verified November 2026
Anchor: #eu-sheffield-nlp
University of Surrey – Institute for People-Centred AI & CVSSP
Media, creative-industry, health and security companies needing computer-vision, sketch/visual search, audio or biometrics R&D partners, and recruiters seeking vision PhDs in the UK.
Medium confidenceDetailsClose
Surrey's AI strength centres on the Centre for Vision, Speech and Signal Processing (CVSSP; 150+ researchers, ~30 years of audio-visual perception work) and its sister Institute for People-Centred AI, both directed by Adrian Hilton. Surrey ranks #3 in Europe for computer-vision output in CSRankings 2021–2025.
Strengths
- #3 in Europe for computer-vision venue output in CSRankings 2021–2025
- CVSSP is one of the UK's largest audio and vision groups (150+ researchers) with spin-outs in biometrics, communications, medical and creative industries
- Eleven Surrey AI & Image Processing authors visible in the 2026 Stanford–Elsevier list, incl. Kittler (#22 in UK ICT)
- People-Centred AI institute links vision research to NHS and BBC-type applications
Watch-outs
- ML-core (#24) and NLP (#46) output modest relative to vision
- Founding dates, funding and named industry partners not given on the pages reviewed
- Not an ELLIS Unit site; institute-level claims ('number one for computer vision') are self-reported
Leading researchers
- Director, CVSSP and founding Director, Institute for People-Centred AI3D vision and 4D performance capture
- Josef KittlerTop 2%Faculty, CVSSPin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Surrey; AI & Image Processing; UK ICT #22). Pattern recognition and biometrics
- Yi-Zhe SongTop 2%Faculty, CVSSPin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Surrey; AI & Image Processing). Sketch-based vision; highest Surrey vision output in CSRankings 2021-25
- Tao XiangTop 2%Faculty, CVSSPin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Surrey; AI & Image Processing (site also shows a second 'Xiang, Tao A.' profile)). Visual recognition and person re-identification
- Xiatian ZhuTop 2%Facultyin the V9 (Aug 2026) list per the topcitedscientists.com mirror (University of Surrey; AI & Image Processing). Efficient vision and representation learning; top Surrey ML-core output in CSRankings 2021-25
- Website
- surrey.ac.uk
- Kind
- University group / lab
- Institution
- University of Surrey
- Location
- Guildford, United Kingdom
- CSRankings signal
- #24 in Europe for ML core (ICML/NeurIPS/ICLR; 16.4 adj., 17 faculty) but #3 for vision (CVPR/ECCV/ICCV; 43.0 adj.); #10 combined, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Computer vision
- Efficient ML
- Trustworthy / safe AI
How to engage
- PhD
- Industry partnership programme
- Spin-out / tech transfer
- Visiting researcher
Sources · verified November 2026
Anchor: #eu-surrey-pai-cvssp
University of Trento DISI (MHUG & ELLIS Unit Trento, with FBK)
Partners in generative video/image animation, human-behaviour and multimodal understanding, remote sensing and Earth observation, plus PhD recruitment from a 250+ student doctoral school.
Medium confidenceDetailsClose
The Department of Information Engineering and Computer Science (DISI) hosts a strong vision and multimodal-learning cluster (Sebe's MHUG, a joint vision-and-learning lab with Fondazione Bruno Kessler) and co-hosts the ELLIS Unit Trento with FBK, established March 2023.
Strengths
- Top-15 European vision output (#13 vision, CSRankings 2021–2025) from a mid-sized department
- Widely reused generative-animation code (First Order Motion Model, ~15k GitHub stars)
- Close coupling with FBK research institute via joint labs and the shared ELLIS unit (9 Fellows and Scholars listed)
- Distinctive remote-sensing and planetary-radar expertise (ESA JUICE RIME) alongside core ML
Watch-outs
- Mid-table ML-core output (#38 in Europe) relative to its vision strength
- Small city and modest department size limit breadth in LLM-scale or robotics research
- Department-level funding and headcount not published on the pages reviewed
Leading researchers
- Nicu SebeTop 2%Professor, DISI; leader of the Multimedia and Human Understanding Group (MHUG); ELLIS Unit Trento co-director (with Bruno Lepri, FBK)listed as 'Sebe, Niculae' on the topcitedscientists.com Università di Trento page (AI & Image Processing). Top Trento ML/vision output in CSRankings
- Elisa RicciTop 2%Associate Professor, DISI, and senior researcher, FBK; scientific manager of the FBK–DISI joint Vision and Learning lab; ELLIS Fellowlisted on the topcitedscientists.com Università di Trento page (AI & Image Processing). Domain adaptation, human behaviour analysis
- Lorenzo BruzzoneTop 2%Professor, DISI; contact for the Remote Sensing Laboratory; ELLIS Fellowlisted on the topcitedscientists.com Università di Trento page (Geological & Geomatics Engineering). Radar sounder for ESA's JUICE mission (RIME)
- Kind
- University group / lab
- Institution
- University of Trento
- Location
- Trento, Italy
- CSRankings signal
- #38 in Europe for ML core (9.3 adj.), #13 for vision, #23 for NLP, 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Computer vision
- NLP / LLMs
- Trustworthy / safe AI
- ML for science
How to engage
- PhD (IECS Doctoral School) / ELLIS PhD programme
- Joint labs with Fondazione Bruno Kessler (FBK)
- Industry and public-administration collaboration; spin-offs
- MSc in Artificial Intelligence Systems graduate hiring
Sources · verified November 2026
Anchor: #eu-trento-disi-ellis-trento
University of Warsaw – Faculty of Mathematics, Informatics and Mechanics (MIMUW): AI & ML
Partners in explainable and responsible ML, algorithmic game theory/social choice and computational biology, and recruiters seeking strong algorithms/ML graduates in Central Europe. For state, defence or public-sector applied AI see the IDEAS Research Institute entry.
Medium confidenceDetailsClose
The University of Warsaw's Faculty of Mathematics, Informatics and Mechanics is strong in algorithms, classic AI (AAAI/IJCAI) and explainable ML (Biecek's MI².AI). Since Dec 2025 UW has co-hosted ELLIS Unit Warsaw (est. Nov 2023), which is hosted by the state IDEAS Research Institute and directed by Tomasz Trzciński (IDEAS/WUT); IDEAS is listed separately (see ideas-research-institute).
Strengths
- Strong classic-AI and algorithms output: University of Warsaw #19 in Europe for AAAI/IJCAI (CSRankings 2021–2025)
- Widely used open-source explainable-AI tooling (DALEX, fairmodels, survex) from Biecek's MI².AI group
- Co-host (since Dec 2025) of Poland's first ELLIS unit, which is hosted by IDEAS Research Institute (separate entry) with WUT
- Deep algorithms tradition at MIMUW supplies strong ML engineering talent
Watch-outs
- Low ML-core venue output (#75 in Europe, CSRankings 2021–2025)
- ELLIS activity is hosted outside UW (by IDEAS RI), and UW's AI groups have no single institute or website
- Faculty website gives little detail on AI groups, size or funding; Stanford–Elsevier status of UW researchers not checked for this entry
Leading researchers
- Professor, University of Warsaw; Director, IDEAS Research Institute; ELLIS FellowAlgorithms and ML; former president of IDEAS NCBR, chosen to lead the new IDEAS institute from Jan 2025
- Full Professor, University of Warsaw (MIMUW); Director, Centre for Credible AI, Warsaw University of Technology; head of MI².AI; ELLIS Fellow (listed under WUT)Explainable ML ('model science'); creator of DALEX and the DrWhy.AI toolset
- University of Warsaw; also Helmholtz Munich (per ELLIS); ELLIS Unit Warsaw memberGenerative models for antimicrobial peptides, tumour evolution and digital pathology
- Website
- mimuw.edu.pl
- Kind
- University group / lab
- Institution
- University of Warsaw (with Warsaw University of Technology and IDEAS Research Institute)
- Location
- Warsaw, Poland
- CSRankings signal
- University of Warsaw: #75 in Europe for ML core (4.1 adj.), #19 for AI (AAAI/IJCAI), 2021–2025 (CSRankings data)
Flagship outputs
Focus areas
- Trustworthy / safe AI
- Theory
- ML for science
- RL
How to engage
- PhD / ELLIS PhD programme
- Applied projects with IDEAS Research Institute (state AI institute)
- Spin-out / tech transfer (e.g. MI².AI-linked start-ups)
- Graduate hiring from MIMUW
Sources · verified November 2026
Anchor: #eu-warsaw-uw-ellis-warsaw
Villum Foundation (Villum Fonden) – data-science and AI funding (Villum Synergy, DDSA, P1)
Permanent staff at Danish universities pairing computer science, statistics or applied maths with another discipline (not health); researchers seeking large open-topic personal grants in Denmark.
Medium confidenceDetailsClose
Danish foundation (founded 1971 by Villum Kann Rasmussen; ~DKK 1.82bn distributed in 2025) whose AI-relevant funding centres on Villum Synergy grants pairing computer science with other fields, plus co-funding of DDSA and P1.
Strengths
- Dedicated, recurring programme for computer-science-plus-domain research (Villum Synergy, since 2019)
- Sizeable multi-year grants: full projects DKK 9–14m over 3–5 years, plus an indirect-cost supplement
- Co-funds national AI and data-science capacity: Pioneer Centre for AI (P1) and the Danish Data Science Academy
- Funded projects are ML-heavy: explainable AI, generative AI, physics-informed and protein-fitness ML
Watch-outs
- Restricted to permanent staff at Danish universities; health science excluded from Villum Synergy
- No standalone AI strategy or AI budget; AI funding is spread across open-topic programmes
- DDSA's 5-year budget ends Dec 2026 with no successor announced on its site; no Villum Synergy call open or announced as of Nov 2026
Leading researchers
- Executive Director, Villum Fonden (since Aug 2024)Previously Director General of the Danish Medicines Agency
- Head of Programme, Villum SynergyNamed programme contact for Villum Synergy applications
- Professor, DTU Compute; collaborator on Villum Synergy 2025 project 'Orienteering' (lead: Carlos Acevedo-Rocha, DTU Biosustain; DKK 4.5m)Epistasis-aware ML for protein fitness landscapes, with Carlos Acevedo-Rocha (DTU Biosustain)
- PI, Villum Synergy 2024 project AInterviewer (IT University of Copenhagen)Generative AI for large-scale qualitative interview data, with Hjalmar Carlsen (University of Copenhagen)
- Website
- villumfonden.dk
- Kind
- Foundation / funder
- Institution
- Part of the VELUX Foundations (with Velux Fonden)
- Location
- Søborg (Copenhagen area), Denmark
- Founded
- 1971
- Funding scale
- ~DKK 1.82bn (≈€244m) distributed in 2025 across all areas; Villum Synergy DKK 60m per call; DKK 31.8m to DDSA (2021–26); co-funder of P1 (DKK 352m over 13 years, five foundations)
Flagship outputs
- Villum Synergy: annual call (DKK 60m in the latest call); initiation projects from DKK 4.5m, full projects DKK 9–14m; no open or upcoming call listed as of Nov 2026
- Villum Synergy 2024 awards: DKK 47m to 10 projects incl. explainable AI and generative-AI interview research
- Danish Data Science Academy: PhD/postdoc fellowships, co-funded with Novo Nordisk Foundation (Villum DKK 31.8m)
Focus areas
- ML for science
- Trustworthy / safe AI
- NLP / LLMs
How to engage
- Grants / calls (Villum Synergy; two permanent PIs at Danish universities, one from CS/stats/applied maths)
- Villum Investigator, Young Investigator and Villum Experiment (open-topic personal grants)
- DDSA PhD and postdoc fellowships (2026 calls closed; DDSA's 5-year budget ends Dec 2026)
- Partner on Villum-funded projects at Danish universities
Sources · verified November 2026
Anchor: #eu-villum-foundation
WASP – Wallenberg AI, Autonomous Systems and Software Program
Companies wanting to co-supervise industrial PhDs in Sweden, recruit AI/autonomy graduates, or join a 60-company industry partner network.
Medium confidenceDetailsClose
Sweden's largest individual research programme (SEK 7.2bn total, with Wallenberg Foundation funding secured through 2034), funding AI, autonomous systems and software research and a graduate school with ~500 active PhD students.
Strengths
- Very large, long-term private funding: Knut and Alice Wallenberg Foundation support extended through 2034 (programme total SEK 7.2bn)
- Graduate school at national scale: ~500 active PhD students
- Industry integration: 60 partner companies (as of Jan 2023) and industrial PhD tracks
- Spans five major universities (Chalmers, Linköping, Lund, KTH, Umeå)
Watch-outs
- Swedish partner universities rank modestly in CSRankings ML-core output (KTH #44 in Europe)
- Programme rather than lab: research quality varies by host group
- Primarily for Swedish-based researchers and companies with Swedish presence
Leading researchers
- Program Director (since Jan 2025)AI and robotics researcher (Örebro University); leads WASP
- Chair of the WASP board (Program Director July 2020–Dec 2024)Scientific visualisation; Linköping University
- Danica KragicTop 2%Professor, KTH (WASP partner)in the V9 (Aug 2026) list per the topcitedscientists.com mirror (KTH; subfield Industrial Engineering & Automation). Robotics and robot learning
- Website
- wasp-sweden.org
- Kind
- Foundation / funder
- Institution
- Funded by the Knut and Alice Wallenberg Foundation
- Location
- Chalmers, Linköping, Lund, KTH and Umeå universities, Sweden
- Founded
- 2015
- CSRankings signal
- Partner KTH: #44 in Europe for ML core (8.1 adj.), 2021–2025 (CSRankings data)
- Funding scale
- Programme total SEK 7.2bn; Knut and Alice Wallenberg Foundation support secured through 2034 (Sept 2026); earlier figures: SEK 6.2bn total incl. SEK 4.9bn from the Foundation
Flagship outputs
Focus areas
- Robotics learning
- Efficient ML
- Theory
- Trustworthy / safe AI
How to engage
- Industrial PhD / co-supervision
- Industry partnership programme
- Grants / calls (recruitment packages)
- Graduate school events
Sources · verified November 2026
Anchor: #eu-wasp-wallenberg
Zuse Institute Berlin (ZIB) – AI in Society, Science and Technology (AIS²T) & IOL Lab
Researchers and companies working on efficient deep learning, ML–optimisation hybrids or AI for maths and science, and PhD/postdoc candidates wanting a Berlin maths base (MATH+, Berlin Mathematical School).
Medium confidenceDetailsClose
Berlin's applied-maths and HPC institute; its AIS²T department and Sebastian Pokutta's IOL Lab combine optimisation and ML (efficient/sparse deep learning, AI4Science, AI4Math), with regular ICML/NeurIPS/ICLR papers.
Strengths
- Regular output at top ML venues (ICML, NeurIPS, ICLR) from a small lab, centred on efficient and sparse deep learning
- Rare ML–optimisation depth: Frank-Wolfe theory, mixed-integer solvers (SCIP, Boscia.jl) and learning combined
- Own HPC: NHR national centre with a new GPU system for AI and on-premise LLM deployment
- Embedded in MATH+, renewed in 2025 with an explicit focus on the mathematical foundations of AI
Watch-outs
- ML is concentrated in one department; most of ZIB works on optimisation, simulation and HPC
- No structured ML PhD programme; PhD/postdoc funding is case by case and visits are unfunded
- Budget and staff figures are not published; much ML output is credited to TU Berlin rather than ZIB
Leading researchers
- Vice President, ZIB (with Andrea Walther); head of the IOL Lab (ZIB AIS²T and TU Berlin)Frank-Wolfe methods, sparse/compressed networks; #44 European author by ICML/NeurIPS/ICLR output 2021–2025 (CSRankings)
- President, ZIBLeads the institute; ZIB co-supports the renewed MATH+ Cluster of Excellence
- Area lead, iol.LEARN (deep learning), IOL LabEfficient deep learning (pruning, quantisation) and AI4Forest; co-author of Sparse Model Soups (ICLR 2024)
- Deputy lab head, IOL LabNeural discovery in mathematics (ICML 2025)
- Website
- iol.zib.de
- Kind
- Research institute
- Institution
- Zuse Institute Berlin (non-university research institute of the State of Berlin); IOL Lab is joint with TU Berlin
- Location
- Berlin, Germany
- Founded
- 1984
- CSRankings signal
- Not listed separately in CSRankings; IOL head Sebastian Pokutta (listed under TU Berlin) is #44 among European authors for ML core (6.3 adj.), 2021–2025
- Funding scale
- State of Berlin institute; total budget not stated in the 2025 annual report. MODAL phase 3 (from Apr 2025) up to €8m from BMFTR; MATH+ renewed for 7 years (May 2025)
Flagship outputs
- AI4Forest global 10 m canopy height map (ICML 2024 'Estimating Canopy Height at Scale'), code on GitHub
- FrankWolfe.jl: MIT-licensed Julia library for Frank-Wolfe / conditional-gradient methods (~117 stars)
- iol.LEARN publication list: e.g., 'A Free Lunch in LLM Compression' (NeurIPS 2026), Sparse Model Soups (ICLR 2024)
Focus areas
- Efficient ML
- Theory
- ML for science
- Trustworthy / safe AI
How to engage
- PhD (via Berlin universities / Berlin Mathematical School; funding case by case from MATH+, MODAL, DFG)
- Postdoctoral research (rolling applications by research area)
- Visiting researcher (2–6 month unfunded student research visits)
- Industry partnership via the MODAL research campus (17 industry partners)
- Compute: ZIB runs an NHR national HPC centre (Lise) and a new GPU system for AI workloads
Sources · verified November 2026
Anchor: #eu-zib-ais2t-iol
Nothing matches these filters.
How this list is built
Researchers drew each entry from public evidence. A second person who did not write it then opened every link, checked current roles and traced each strength and watch-out to a source. Nobody pays to be listed.
- Stanford–Elsevier top-2% scientists list, 2026 edition (Version 9). Each badge was checked researcher by researcher against a public mirror of the dataset. No badge means not confirmed, not absent.
- CSRankings publication counts for European institutions in machine learning, AI, vision and NLP.
- ELLIS units, ERC grants and each group’s own outputs. Open calls are dated; check the group’s site before applying.
Spotted an error or a missing group? Email us with a link and we update the entry.
Europe’s ML output leaders
Papers at ICML, NeurIPS and ICLR, 2021–2025, with each paper split across its authors’ institutions. Source: CSRankings data, November 2026. Bars are to scale.
Bring us a question worth checking.
Corrections and additions
Send a correctionA wrong fact, a missing group or firm, or a newer report. Send the link and we update the entry.
Research collaborations
Discuss researchThreat models, benchmarks and private-inference prototypes.