Prof. Dr. Thomas Hofmann is a Full Professor and Head of the Department of Computer Science at ETH Zurich since 2014. He also leads the Institute for Machine Learning. His research focuses on machine learning, deep learning, natural language understanding, and text understanding. Hofmann holds a Ph.D. from the University of Bonn (1997) and has held academic positions at Brown University (1999–2004) and TU Darmstadt. He transitioned to industry as Director of Engineering at Google (2006–2014), leading the Zurich R&D center, before returning to academia. He co-founded Recommind (2000) and 1plusX (Swiss marketing tech company), currently serving as Chief Scientist and board member at 1plusX. Education: Ph.D. in Computer Science, University of Bonn (1997) Postdoctoral Work: MIT (CBCL/AI Lab), UC Berkeley (EECS/ICSI) His research explores advanced machine learning techniques, including diffusion models, generative adversarial networks, and optimization dynamics. Hofmann’s entrepreneurial ventures reflect his focus on applying AI to real-world challenges, such as e-discovery and marketing technology. His work spans theoretical contributions (e.g., neural network training dynamics, continual learning) and applied innovations (e.g., image editing, portrait generation). Hofmann actively bridges academia and industry, influencing both research and commercial AI applications.
Fanny Yang is an Assistant Professor in the Computer Science Department at ETH Zurich . She previously held postdoctoral positions at Stanford University and a Junior Fellowship at the Institute for Theoretical Studies at ETH Zurich, advised by Nicolai Meinshausen . Her PhD was completed at the EECS Department of UC Berkeley , supervised by Martin Wainwright . Research Interests : Theoretical foundations of machine learning and statistics , particularly focusing on overparameterized models and high-dimensional data . Developing trustworthy ML models with emphasis on distributional robustness , domain generalization , and interpretability . Applications in medical diagnostics and treatment effect analysis , aiming to address reliability issues in real-world domains. Recent Work Trends : 2025 Articles : Theoretical analysis of semi-supervised multi-objective learning , test-time scaling with verifiers, and foundation models for efficient randomized experiments . 2024 Articles : Studies on robust mixture learning , privacy-preserving data synthesis via optimal transport , confounding quantification in causal inference , and semi-private learning frameworks. 2023 Articles : Investigations into active vs. passive learning in high dimensions, inductive bias in noisy interpolation , and semi-supervised novelty detection using model ensembles .
Liza Rebrova is an Assistant Professor in the Department of Operations Research and Financial Engineering (ORFE) at Princeton University's School of Engineering and Applied Science. She is also an associated faculty member of the Program in Applied and Computational Mathematics (PACM) and the Center for Statistics and Machine Learning (CSML). Her research focuses on randomized numerical linear algebra and the mathematics of data science, with connections to high-dimensional probability and stochastic optimization. She develops mathematically justified randomized algorithms for large-scale data, particularly interested in settings where data exhibits mathematical structure such as spectral decay, multi-modality, or non-negativity. Her work addresses challenges in data compression, recovery, and optimization under structured noise conditions. Dr. Rebrova's recent publications show a strong trend toward developing robust algorithms for linear systems, tensor data compression, and nonnegative matrix factorization. Her work bridges theoretical foundations with practical applications in machine learning and data science, with emphasis on memory efficiency and handling corrupted or incomplete data. NSF DMS-2309685 (Single PI, 2024-2026): "Outliers are not what they seem: data-aware, flexible, and robust randomized iterative methods" NSF DMS-2108479 (Collaborative, 2022-2024): "Fast, Low-Memory Embeddings for Tensor Data with Applications" (co-PIs Mark Iwen and Deanna Needell) She currently advises three PhD students (Jackie Lok, Shambhavi Suryanarayanan, and Sofiia Shvaiko) and has previously advised Abraar Chaudhry (PhD 2024) and Nicolo Grometto (Masters thesis 2023). At Princeton, she teaches graduate courses including ORF526 (Graduate Probability), ORF387 (Networks), and ORF523 (Convex and Conic Optimization).
Mathieu Salzmann is a Senior Scientist and Lecturer at École Polytechnique Fédérale de Lausanne (EPFL), affiliated with the Computer Vision Laboratory (CVLAB) in the School of Computer and Communication Sciences (IC). He also holds a courtesy appointment with the EPFL College of Humanities and serves as Deputy Chief Data Scientist at the Swiss Data Science Center (SDSC). He has held concurrent roles in teaching units including SIN, SODH, and SSC, reflecting his interdisciplinary engagement. His research focuses on the intersection of machine learning and computer vision, particularly in deep learning for 2D and 3D visual scene understanding, efficient and robust models, domain adaptation, and interpretable AI. These interests are evident across his extensive publication record in top-tier venues. His recent publications (2023–2024) show a consistent trend in advancing deep learning methods for visual recognition, with strong representation at CVPR, ICCV, ECCV, ICML, ICLR, and NeurIPS. Topics include domain generalization, 3D understanding, model robustness, and multimodal learning, often with applications in real-world systems. His editorial roles as Associate Editor for IEEE TPAMI and Action Editor for TMLR further highlight his leadership in the field. Area Chair: ICML 2023, CVPR 2023, ICCV 2023, NeurIPS 2023, AAAI 2024, ECCV 2024 Associate Editor: IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) Action Editor: Transactions on Machine Learning Research (TMLR) Mathieu Salzmann has supervised numerous PhD students at EPFL, both current and past, including Bouquet Yann Yanis, Javed Saqib, Li Shuangqi, and others. He has also been involved in research grants and collaborative projects, such as his work with S. Süsstrunk and R. Baroni on comics reconfiguration. His part-time role as Senior GNC Engineer at ClearSpace (2020–2024) illustrates his applied research engagement in aerospace systems. He is actively involved in EPFL’s data science and AI research ecosystem through SDSC and multiple labs.
Caglar Gulcehre is a Tenure Track Assistant Professor at École Polytechnique Fédérale de Lausanne (EPFL) and Principal Investigator of the CLAIRE (Caglar Gulcehre Laboratory of Artificial Intelligence Research) lab. Previously, he worked as a Staff Research Scientist at Google DeepMind, Microsoft Research, and IBM Research. His research focuses on reinforcement learning , foundation models , LLM alignment , and sequence modeling . Current Position : Assistant Professor, EPFL Lab : CLAIRE Lab Previous Roles : Staff Research Scientist at DeepMind, MSR, IBM Research His work spans reinforcement learning , deep learning , and neural architecture design , with a focus on safety , trustworthy AI , and real-world applications . He has published in top venues including Nature , NeurIPS , ICML , and JMLR . Scientific contributions include: Best paper award at NeurIPS Nonconvex Optimization workshop Honorable mention for best paper at ICML 2019 Co-organizer of seven workshops at NeurIPS, ICML, and ICLR He supervises PhD students in areas related to AI for algorithm discovery , neural architectures , and foundation models , including: Skander Moalla Justin Samuel Deschenaux Liangze Jiang Xiuying Wei Yitao Xu
Maksym Andriushchenko is an incoming faculty member at the ELLIS Institute Tübingen and Max Planck Institute for Intelligent Systems , where he will lead the AI Safety and Alignment group as a principal investigator. Currently, he is a Doctoral Assistant at the Theory of Machine Learning Laboratory (TML) within the Department of Computer Science (IINFCOM) at Swiss Federal Institute of Technology in Lausanne (EPFL) . His work focuses on AI safety, adversarial robustness, and alignment of large language models (LLMs) with societal values.
Mikaela Iacobelli is an Associate Professor in the Department of Mathematics at ETH Zürich. During the 2024-25 academic year, she was a von Neumann Fellow at the Institute for Advanced Study in Princeton. Previously, she held faculty positions at Durham University and a research fellowship at the University of Cambridge. Her educational background includes: PhD in Mathematics from Sapienza University of Rome and École Polytechnique in Paris (2015) Master's degree from Sapienza University of Rome (2012) Bachelor's degree from Sapienza University of Rome (2009) Mikaela's research lies at the interface of analysis, kinetic theory, and statistical mechanics. She studies partial differential equations that model the collective behavior of many-particle systems, with a focus on Vlasov-type plasmas and gravitational dynamics. Her current projects range from quasineutral and singular-limit problems for Vlasov-type systems to quantization of measures, ultrafast diffusion, and gradient-flow structures that link microscopic particle models to macroscopic fluid descriptions. She makes extensive use of PDEs techniques, optimal transport, probability, calculus of variations, and Riemannian geometry in her work. Her recent publications demonstrate a strong focus on Vlasov-type equations, particularly examining quasineutral limits, stability properties, and connections to other physical systems like Euler equations and magnetohydrodynamics. She has made significant contributions to understanding Landau damping, quantization problems on manifolds, and the mathematical foundations of plasma physics. Her notable scientific awards include: SNSF Starting Grant (Swiss ERC) Challenges and Breakthroughs in the Mathematics of Plasmas (2025-2030) von Neumann Fellow at the Institute for Advanced Study, Princeton (2024-2025) Invited speaker at the International Congress of Mathematical Physics (2021) CO-PI of the Germaine de Staël Funding Program for French-Swiss cooperation (2021-2023) L'Oréal prize for Women in Science (2015) Mikaela actively mentors postdocs, PhD, Master's, and Bachelor's students. Her current mentees include postdocs Dennis Chemnitz, Rishabh Gvalani, Stefano Rossi, and Simon Becker, as well as PhD students Thérèse Moerschell, Ata Deniz Aydin, and Antoine Gagnebin. She has served as PI for the Starting Research Grant from the University of Rome Sapienza and is currently the PI for the SNSF Starting Grant. She co-organizes several academic seminars including the Zurich Colloquium in Mathematics, the PDE and Mathematical Physics seminar at ETH Zürich and UZH, and the Analysis Seminar. She also serves on various committees including as Chair of the European Mathematical Society Committee for Women in Mathematics.
Giancarlo Ferrari Trecate is an Adjunct Professor at the Swiss Federal Institute of Technology Lausanne (EPFL) , affiliated with the School of Engineering and the SCI-STI-GFT department. He is also involved in teaching and research through the STI-SGM and EDRS-ENS programs. Research Interests : Automatic control, state estimation, system identification, machine learning, distributed control, hybrid systems, microgrids, biochemical networks, voltage and frequency stabilization in AC/DC microgrids. Publications Trends : His recent work focuses on integrating Neural ODEs and Hamiltonian structures for stable control systems, regret minimization in distributed control, and robust state estimation under uncertainty. Applications include autonomous mobility-on-demand , power grid optimization , and secure microgrid control against cyber-attacks. Scientific Awards : No specific awards mentioned in the provided data. Teaching & Advising : He supervises PhD students in mechanical engineering and teaches courses on Multivariable control and Networked control systems . His lab, DECODE , specializes in Dependable Control and Decision systems.
El Mahdi Chayti is a doctoral researcher at the École Polytechnique Fédérale de Lausanne (EPFL), affiliated with the School of Computer and Communication Sciences and the Department of Computer Science. He works under the Machine Learning and Optimization (MLO) lab, focusing on advanced optimization techniques and machine learning theory. Contact: el-mahdi.chayti@epfl.ch . Current Roles: Doctoral Assistant and PhD Student in Computer and Communication Sciences Research Interests: Machine Learning, Optimization Algorithms, Meta-learning, Second-order Optimization, Energy Forecasting His publications highlight expertise in stochastic and cubic Newton methods, hybrid deep learning models, and personalized collaborative learning. Recent works include Improving Stochastic Cubic Newton with Momentum (2025) and Hybridization of Deep Learning with Physical Knowledge for Energy Forecasting (2019) . Key trends in his research span optimization efficiency , meta-learning frameworks , and integration of domain knowledge into AI models. He emphasizes theoretical guarantees and practical scalability in algorithm design.
Zsolt Patakfalvi is an Associate Professor at École Polytechnique Fédérale de Lausanne (EPFL), holding positions in the School of Basic Sciences (SB) within the Department of Mathematics (MATH). He is affiliated with the Chair of Algebraic Geometry (CAG) and the Section of Mathematics for Engineers (SMA-ENS). Additionally, he serves as Director of SMA-GE and holds roles in academic governance bodies like the Conference of Section Directors (CDS) and SB Faculty Management. His research focuses on Algebraic Geometry, particularly in birational geometry, positive characteristic methods, moduli theory, and mixed characteristic algebra. He explores topics such as Hodge theory, singularities, and applications to arithmetic geometry. Notable contributions include work on the minimal model program, test ideals, and counterexamples to classical conjectures in positive characteristics. He supervises doctoral students in areas like algebraic geometry and commutative algebra, including Jefferson Baudin, Léo Navarro Chafloque, and Linus Rösler. His past advisees include Emelie Arvidsson and Quentin Posva. Patakfalvi’s publications frequently address foundational questions in geometry, with recent work extending into perfectoid spaces and globally-regular varieties. He coordinates courses such as 'Algebra III - Rings and Fields' and 'Perfectoid spaces' at EPFL, reflecting his commitment to both research and education. His academic service includes managing educational programs within SB-SMA and contributing to institutional decision-making through CDS membership.
Rima Alaifari is currently an Assistant Professor for Applied Mathematics at ETH Zürich , where she works on applied analysis, inverse problems, and scientific machine learning. Her research emphasizes stability analysis and regularization of inverse problems, applied harmonic analysis, phase retrieval, and operator learning. She is an associated member of the ETH AI Center and will transition to a full professorship at RWTH Aachen University in 2025 as Chair of Analysis and its Applications. Education : PhD in Mathematics (2010–2014, Vrije Universiteit Brussel); MSc in Applied and Industrial Mathematics (2005–2010, Johannes Kepler University) Research Focus : Stability estimates for inverse problems, phase retrieval in wavelet/Gabor transforms, operator learning with neural networks, and deep learning robustness. Article Trends : Her recent work bridges harmonic analysis with machine learning, focusing on phase retrieval stability, adversarial perturbations in imaging, and mathematically grounded neural operator frameworks like ReNO and CNO. Advising : She has supervised PhD students like Tandri Gauksson and Matthias Wellershoff. Former postdoctoral researchers include Francesca Bartolucci (now at TU Delft) and Jesse Railo (Finnish Inverse Prize winner).
Francesca Da Lio is a Professor at the Department of Mathematics, ETH Zurich, where she has held a titular professorship since 2014. Her research focuses on nonlinear elliptic and parabolic partial differential equations (PDEs), with applications in stochastic and deterministic optimal control, homogenization, front propagation, and geometric analysis. She has pioneered work on conformally invariant variational problems and nonlocal PDEs, including fractional harmonic maps and stability analysis for critical points. PhD in Mathematics (1998) and Summa Cum Laude Degree in Mathematics (1994) from University of Padova. Her research explores the interplay between nonlinearity and non-locality, particularly in problems arising from geometry, mathematical finance, and physics. She has led major Swiss National Fund (SNF) projects, including grants for geometric analysis and conformally invariant variational theory. Her work on 3-commutators, integrability by compensation, and Morse index stability has advanced the understanding of harmonic maps and elliptic systems. Francesca Da Lio has mentored numerous PhD, postdoctoral, and Master/Bachelor students, including Dominik Schlagenhauf, Jerome Wettstein, and Ali Hyder. She has served on hiring committees for full professorships at ETH Zurich and co-organized international conferences such as 'Recent Advances in Nonlocal and Nonlinear Analysis' and 'Topics in Sub-Elliptic PDEs.' Scientific Awards: Italian Scientific Qualification as Full Professor in Mathematical Analysis (2013). She contributes to editorial boards, including Advances in Calculus of Variations , and participates in academic services like refereeing for SNF projects and international journals.
Chen Liu is an Assistant Professor in the Department of Computer Science at City University of Hong Kong and the Principal Investigator (PI) of the Machine Learning and Optimization (MLO) group. His research focuses on building reliable machine learning models, particularly studying robustness and privacy properties of deep neural networks from an optimization perspective. University: City University of Hong Kong Academic Rank: Assistant Professor Students: Supervises multiple PhD, MPhil, and postdoctoral researchers. Education: Holds a Ph.D. (2022) and MSc (2017) in Computer Science from École Polytechnique Fédérale de Lausanne (EPFL), and a BSc (2015) in Computer Science from Tsinghua University. Research Interests: Adversarial robustness, privacy-preserving machine learning, optimization algorithms, dataset distillation, generative models, and theoretical analysis of loss landscapes. His work addresses challenges like catastrophic overfitting, architecture overfitting in distilled data, and stable adversarial training methods. Article Trends: Recent publications explore adversarial robustness under l0/l1 norms, gradient inversion for data reconstruction, evolutionary factor searching in finance, and meta-tuning for out-of-domain few-shot learning. These works emphasize optimization techniques to enhance model reliability and generalization. Scientific Awards: Microsoft Research Ph.D. Scholarship Programme (2017–2019) Advising and Grants: Supervises a diverse team of current and former students, with collaborations across institutions like George Mason University and Zhejiang University. Research supported by academic and industry grants. Labs and Teams: Leads the MLO group, which investigates fundamental ML theory and algorithms to improve system reliability. The group's work spans adversarial training, dataset distillation, and generative model optimization.
Desmond Elliott is an Associate Professor and Villum Young Investigator at the Department of Computer Science, University of Copenhagen. His research focuses on vision-language models, multilingual and multimodal processing, with particular emphasis on tokenization-free language modeling approaches. He leads a research group actively working on pixel language models and cross-lingual multimodal understanding. University of Copenhagen, Department of Computer Science Villum Young Investigator Associate Editor for JAIR (2025-2028) Senior Area Chair for ACL 2025 Elliott's research spans vision-language integration, multilingual NLP, and multimodal machine learning. His work explores how language models can operate directly on visual pixels without traditional tokenization, enabling more seamless integration of vision and language processing. He investigates compositional generalization in multimodal systems, retrieval-augmented image captioning, and cross-lingual transfer in vision-language tasks. His group develops methods for low-resource language processing and creates benchmarks for evaluating multimodal systems across diverse cultural contexts. His recent publications demonstrate strong trends in pixel-based language modeling, synthetic dataset generation through retrieval augmentation, and multilingual vision-language processing. The work spans theoretical advances in model architectures and practical applications in areas like medical text analysis, food culture understanding, and social media content moderation. His research often bridges computer vision and natural language processing with a focus on making these technologies accessible across diverse languages and cultures. Best Paper Honorable Mention at CVPR Visual Concepts Workshop 2025 Best Long Paper Award at EMNLP 2021 Area Chair Favourite paper at COLING 2018 Elliott actively supervises student projects in BSc and MSc programs related to his research interests. His research has received substantial funding from Google (2024-2025), Facebook (2022-2024), Villum Foundation (2021-2026), Novo Nordisk Foundation (2019-2024), and European Union (2023-2026). He regularly recruits postdocs for projects including the Danish Foundation Models project and the Responsible AI for the People Project. His group holds regular meetings on Tuesdays from 13:00-14:00 in IF G.03, with an active mailing list for announcements. The research environment appears collaborative, with frequent co-authorship across institutions and regular participation in major NLP and computer vision conferences.
Florent Krzakala is a Full Professor at École polytechnique fédérale de Lausanne (EPFL) in Switzerland, holding positions across multiple departments including the School of Basic Sciences (SB), School of Engineering (STI), and specifically within the Department of Physics (IPHYS) and Department of Electrical Engineering (IEM). He leads the Information, Learning and Physics Laboratory (IdePHICS) and maintains an office at ELD 239, Station 11, 1015 Lausanne. His research bridges statistical physics and computational disciplines, with significant contributions to understanding the theoretical foundations of machine learning and optimization problems. Dr. Krzakala received his MSc in Physics from Orsay, France in 1999, followed by a PhD in Statistical Physics from Orsay, Paris XI, France in 2002, and completed a postdoctoral position at Roma La Sapienza in 2004. This strong foundation in physics has informed his interdisciplinary approach to computational problems. His research interests span Statistical Physics, Machine Learning, Probability and Statistics, Computer Science, Information Theory, Inference on Graphs, Random Constraint Optimization, and Computational Optics. Krzakala's work focuses on applying methods from statistical physics to problems in theoretical computer science, probability, and machine learning. He investigates how concepts from disordered systems and phase transitions can illuminate computational barriers in optimization and inference tasks. His research has particular relevance for understanding the behavior of neural networks, compressed sensing, and high-dimensional statistical models. Analysis of his recent publications reveals a strong trend toward understanding the fundamental limits of learning in high-dimensional settings, with particular emphasis on phase transitions, statistical-to-computational gaps, and the theoretical properties of deep learning architectures. His work frequently bridges rigorous mathematical analysis with practical machine learning applications, demonstrating how insights from statistical physics can inform algorithm design and theoretical understanding in AI. Krzakala actively mentors the next generation of researchers, supervising numerous PhD students whose work continues to advance these interdisciplinary fields. His laboratory serves as a hub for researchers exploring the intersection of physics and computation, fostering collaborations across traditional disciplinary boundaries. He teaches advanced courses including Fundamentals of Inference and Learning, Statistical Physics, and Statistical Physics for Optimization & Learning, which examine the connections between physical principles and computational methods. His educational materials, including lecture notes on statistical physics methods in optimization and machine learning, have become valuable resources for students and researchers worldwide. As founder and scientific advisor of the startup Lighton, Krzakala has also demonstrated a commitment to translating theoretical insights into practical applications, particularly in the realm of optical computing for machine learning tasks.
