EPFL AI Center

The EPFL AI Center is the artificial intelligence research center of the École Polytechnique Fédérale de Lausanne (EPFL), the Swiss federal technology institute located in Lausanne, Switzerland. The center was established in 2021 to coordinate AI research across approximately 100 EPFL faculty members spanning the School of Computer and Communication Sciences, the School of Engineering, the School of Basic Sciences, and adjacent EPFL units. The EPFL AI Center is co-directed by Marcel Salathé (digital epidemiology professor and one of the principal European voices on AI policy and digital health) and Pierre Vandergheynst (signal processing and machine learning professor; former EPFL Vice President for Education). The center plays a structurally important role within the broader Swiss AI ecosystem alongside the ETH AI Center at ETH Zurich, with the two institutions jointly anchoring the Swiss AI Initiative — a multi-year SwissNRP program that produced the open-source Swiss-built Apertus large-language-model release in 2024 to 2025. As of April 2026, the EPFL AI Center is one of the principal European academic AI research centers and one of the structurally consequential nodes of the broader European AI research ecosystem alongside ETH AI Center, INRIA, DFKI, Alan Turing Institute, Tübingen AI Center, and Max Planck Institute for Intelligent Systems.

At a glance

• Founded: 2021 in Lausanne, Switzerland, by EPFL to coordinate AI research across the broader EPFL faculty.
• Status: Research center within EPFL, the Swiss federal technology institute. EPFL itself is one of two Swiss federal institutes of technology (alongside ETH Zurich) and is part of the broader ETH Domain governed by the Swiss federal government.
• Funding: EPFL operating budget plus competitive Swiss research-grant funding from the Swiss National Science Foundation (SNF), the Swiss Innovation Agency (Innosuisse), European Research Council (ERC) grants, and adjacent sources. Specific AI Center budget allocation is not separately disclosed.
• Co-Directors: Marcel Salathé and Pierre Vandergheynst. Both are senior EPFL faculty with substantive research output and prior senior administrative roles at EPFL.
• Other notable leadership: Senior EPFL faculty across the principal AI subdisciplines including Martin Jaggi (machine learning), Antoine Bosselut (NLP and reasoning), Caglar Gulcehre (deep learning, formerly Google DeepMind), and adjacent senior researchers. The center maintains research-cooperation relationships with ETH AI Center, Mila, Cohere for AI, and other peer organizations.
• Open weights: Yes, partial. Research outputs released open-source through GitHub and Hugging Face. The Apertus large-language-model line developed under the Swiss AI Initiative cooperation with ETH Zurich was released open-weights through Hugging Face.
• Flagship outputs: Active publication record at NeurIPS, ICML, ICLR, ACL, EMNLP, CVPR, and adjacent major AI venues. Cooperation with ETH AI Center on the Swiss AI Initiative and the Apertus large-language-model release. Cross-disciplinary AI research integration across EPFL's broader research portfolio including AI for science, AI for medicine, AI for robotics, and adjacent applied-AI research areas.

Origins

The EPFL AI Center was established in 2021 by EPFL with the explicit mandate of coordinating AI research across the broader EPFL faculty and serving as the principal point of engagement between EPFL AI research and the broader Swiss-and-international AI research community. The 2021 founding reflected EPFL's strategic decision to position AI as a cross-cutting research priority spanning multiple schools (Computer and Communication Sciences, Engineering, Basic Sciences, Life Sciences, Architecture, Civil and Environmental Engineering) rather than as a research focus contained within a single department.

EPFL itself was established in 1969 as the federal-level upgrade of the École Polytechnique de l'Université de Lausanne, and operates as one of two Swiss federal institutes of technology alongside ETH Zurich. The EPFL faculty had produced substantive AI research output across the 2000s and 2010s — particularly in computer vision, machine learning theory, signal processing, computational neuroscience, and adjacent areas — but the AI research had been distributed across departments without a unifying coordination structure. The 2021 AI Center founding consolidated coordination, provided a unified external-engagement interface, and anchored the broader Swiss AI ecosystem cooperation with ETH Zurich.

The Swiss AI Initiative, announced in 2023 and accelerated through 2024 to 2025, has been the principal joint research-program investment between EPFL AI Center and ETH AI Center. The initiative was structured around the goal of producing a fully open-source Swiss-built large language model, with the resulting Apertus model line released open-weights through Hugging Face in 2024 to 2025. The initiative anchored substantial Swiss federal compute infrastructure (the CSCS Alps supercomputer in Lugano provides the principal compute capacity for the Apertus training program) and senior research cooperation across the two Swiss federal institutes of technology.

The 2024 to 2026 period has continued research output across the EPFL AI Center research-program areas alongside continued Swiss AI Initiative cooperation. Industry coverage has reported continued senior research recruitment at the EPFL AI Center, including hires from frontier-AI labs and adjacent academic peer institutions.

Mission and strategy

The EPFL AI Center's stated mission is to coordinate AI research across the broader EPFL faculty and to anchor EPFL's positioning within the Swiss-and-international AI research community. The strategy combines four threads. First, foundational AI research with publication output at major AI venues across machine learning, computer vision, NLP, robotics, and adjacent core AI subdisciplines. Second, cross-disciplinary AI integration across EPFL's broader research portfolio, including AI for science, AI for medicine, AI for robotics, AI for engineering, and adjacent applied-AI research areas. Third, cooperation with ETH AI Center on the Swiss AI Initiative and the broader Swiss AI ecosystem. Fourth, international research-cooperation with peer European and global academic AI research organizations.

The competitive premise is that European AI research ecosystem requires coordinated multi-institution engagement to compete structurally with US and Chinese AI research programs, and that EPFL's combination of substantial faculty research output, cross-disciplinary breadth, and Swiss-federal-funding stability positions the AI Center as one of the structurally consequential European AI research nodes.

Models and products

• Active academic-publication program. Research publications at NeurIPS, ICML, ICLR, ACL, EMNLP, CVPR, ICCV, and adjacent major AI venues across the EPFL AI Center research-program areas.
• Apertus large-language-model line. Developed under the Swiss AI Initiative cooperation with ETH AI Center. Released open-weights through Hugging Face in 2024 to 2025. Trained on the CSCS Alps supercomputer.
• Cross-disciplinary AI for science research. Including AI applications in biology, medicine, materials science, climate research, and adjacent EPFL research areas.
• Research-cooperation relationships. With ETH AI Center, Mila, Cohere for AI, and adjacent peer organizations.
• Selected open-source contributions. Through the EPFL AI Center GitHub presence and Hugging Face account.

Distribution channels include academic publication, open-weights model and dataset distribution through Hugging Face, and research-cooperation engagement with peer organizations.

Benchmarks and standing

The EPFL AI Center's evaluation framework focuses on academic publication output (paper count and citation impact at major AI venues), the impact of cross-disciplinary AI research on EPFL's broader research portfolio, the visibility of the Apertus open-weights release within the broader research community, and the strength of cooperation with ETH AI Center and peer European AI research organizations.

Industry coverage has consistently characterized the EPFL AI Center as one of the principal European academic AI research centers, with the multi-year cooperation with ETH AI Center and the Apertus open-weights model release as principal validating data points.

Leadership

As of April 2026, the EPFL AI Center's senior leadership includes:

• Marcel Salathé, Co-Director.
• Pierre Vandergheynst, Co-Director.
• Senior EPFL faculty across the principal AI subdisciplines.

The co-director leadership structure reflects the cross-school coordination mandate and provides continuity across multiple EPFL administrative cycles.

Funding and backers

EPFL operating budget plus competitive Swiss research-grant funding from the Swiss National Science Foundation (SNF), the Swiss Innovation Agency (Innosuisse), European Research Council (ERC) grants, and adjacent sources. The Swiss AI Initiative provides additional federal-program funding for the joint EPFL-ETH cooperation.

Industry position

The EPFL AI Center occupies a distinctive position as one of the principal European academic AI research centers, with the substantial EPFL faculty research output, the cross-disciplinary AI integration across EPFL's broader research portfolio, and the cooperation with ETH AI Center on the Swiss AI Initiative. Industry coverage has consistently grouped EPFL AI Center with ETH AI Center, Tübingen AI Center, Max Planck Institute for Intelligent Systems, INRIA, DFKI, and Alan Turing Institute as the principal European academic AI research organizations.

Competitive landscape

• ETH AI Center. The other Swiss federal institute of technology AI center. Direct cooperation partner on the Swiss AI Initiative and the Apertus model release. Cooperation more than competition.
• INRIA, DFKI, Alan Turing Institute, Tübingen AI Center, Max Planck Institute for Intelligent Systems. European academic AI research peers.
• ELLIS Society. Pan-European ML research network. EPFL AI Center participates as an ELLIS Unit.
• Mila, Vector Institute, Stanford AI Lab, MIT CSAIL, Berkeley BAIR, CMU SCS. International academic AI research peers.
• Hugging Face, Cohere for AI, LAION, BigScience. Open-research peer organizations and partners.
• Mistral AI, Aleph Alpha, Kyutai. European commercial AI peers.

Outlook

• Continued Swiss AI Initiative cooperation with ETH AI Center and Apertus model iteration through 2026 to 2027.
• Continued academic publication output at major AI venues.
• Continued cross-disciplinary AI integration across EPFL's broader research portfolio.
• Continued senior research-talent recruitment from frontier-AI labs and academic peer institutions.
• Continued cooperation with the broader European AI research ecosystem through ELLIS and adjacent multi-institution structures.

Sources

• EPFL AI Center. Center reference.
• EPFL. Parent institution.
• Marcel Salathé EPFL profile. Co-Director reference.
• Pierre Vandergheynst EPFL profile. Co-Director reference.
• Apertus on Hugging Face. Swiss AI Initiative open-weights release.
• ETH AI Center. Cooperation partner on the Swiss AI Initiative.