The ICINAi community
Our steering group brings together researchers across institutions, disciplines, and sectors to help shape ICINAi’s scientific direction and shared infrastructure.
Scientific leadership
Researchers helping guide the consortium’s priorities, programs, and standards.
Northeastern University
David Bau is an assistant professor at Northeastern University and director of the National Deep Inference Fabric. His research studies emergent internal mechanisms in large neural networks across language and vision, including causal tracing, model editing, and shared interpretability infrastructure.
Tel Aviv University
Mor Geva is an assistant professor in the School of Computer Science and AI at Tel Aviv University. Her research opens up the inner workings of large language models to improve their transparency, control, and reasoning.
Technion – Israel Institute of Technology
Yonatan Belinkov is an associate professor at the Technion. He develops probing, causal-analysis, and evaluation methods for understanding what language models represent and how their internal computations produce behavior.
Brown University · Google DeepMind
Ellie Pavlick is an associate professor at Brown University and a research scientist at Google DeepMind. She studies conceptual representations, language, reasoning, and learning, often by comparing the behavior of humans and language models.
Stanford University
Christopher Potts is a professor of linguistics at Stanford University, with a courtesy appointment in computer science. His research connects compositionality, causal abstraction, interpretability, information retrieval, and foundation-model programming.
University of Toronto · Vector Institute · Anthropic
Roger Grosse is a professor at the University of Toronto, a Vector Institute faculty member, and a researcher at Anthropic. His work spans mechanistic interpretability, alignment science, neural-network optimization, and representation learning.
Boston University
Aaron Mueller is an assistant professor of computer science at Boston University. His lab studies how concepts are learned and represented in language models, developing tools for circuit discovery, interpretability, and model control.
Boston University
Naomi Saphra is an assistant professor in Computing & Data Sciences at Boston University. Her research examines language-model training dynamics, how structure emerges during learning, and what those processes reveal about generalization and interpretability.
Weizmann Institute of Science
Tamar Rott Shaham is an assistant professor of computer science and applied mathematics at the Weizmann Institute. She develops automated tools to discover and explain model operations, then uses those insights to control model behavior.
University of Utah
Ana Marasović is an assistant professor in the Kahlert School of Computing at the University of Utah. Her work in natural language processing and human-centered AI develops more faithful explanations and evaluations of language models.
University of Edinburgh · University of Amsterdam
Ivan Titov is a professor at the University of Edinburgh and the University of Amsterdam. He studies trustworthy, robust, interpretable, and controllable language models and leads major European training and research programs in natural language processing.
New York University · Google
Tal Linzen is an associate professor at New York University and a staff research scientist at Google. His work connects language-model evaluation and interpretability with linguistics and cognitive science, especially the study of syntactic structure.
Northeastern University
Byron Wallace is a professor at Northeastern University and director of its BS in Artificial Intelligence program. His research develops machine-learning and natural-language-processing methods for health informatics, with particular attention to faithful explanations, biomedical evidence synthesis, and human–AI systems.
Bar-Ilan University
Yanai Elazar is an assistant professor of computer science and AI at Bar-Ilan University. He builds tools for a science of generative models, including methods for interpretability, data attribution, and rigorous evaluation.
University College London · Parallax
Mario Giulianelli is an associate professor of computational linguistics at University College London and a cofounder of Parallax. He combines behavioral evidence with analysis of internal mechanisms to make AI evaluation more rigorous and explanatory.
Parallax
Gabriele Sarti is a cofounder and technical director at Parallax. He develops white-box methods for context attribution, user modeling, deception detection, and model monitoring, alongside open-source libraries for language-model interpretability.
Transluce
Sarah Schwettmann is cofounder and chief scientist at Transluce, a nonprofit research lab building public infrastructure for responsible AI deployment. Her research reverse-engineers neural networks and develops tools for transparent, auditable, and steerable AI systems.
Simplex · Astera Institute
Eric J. Michaud is a research scientist at Simplex and the Astera Institute. He pursues empirical and conceptual accounts of learned computation, including work on sparse feature circuits, representation geometry, and mechanistic program synthesis.
Yale University
R. Thomas McCoy is an assistant professor of linguistics at Yale University, with a secondary appointment in computer science. He connects AI and cognitive science, using language-model interpretation and evaluation to study both machine intelligence and human language.
Stanford University
Noah D. Goodman is a professor of psychology and computer science at Stanford University. His research uses probabilistic models to study language, cognition, and reasoning, including what language-model behavior reveals about internal representations and thought.
Carnegie Mellon University
Zachary Lipton is the Raj Reddy Associate Professor of Machine Learning at Carnegie Mellon University. His research examines the conceptual foundations of interpretability, robust and adaptive machine learning, natural language processing, and real-world decision systems.
Georgia Tech
Anna Ivanova is an assistant professor of psychology at Georgia Tech, where she leads the Language, Intelligence, and Thought Lab. She combines brain imaging, behavioral studies, and computational modeling to study language and thought in humans and AI.
LMU Munich
Hinrich Schuetze is professor and chair of computational linguistics at LMU Munich. His research spans statistical natural language processing, representation learning, lexical semantics, and the analysis of multilingual language models.
Tel Aviv University
Amir Globerson is a professor of computer science and AI at Tel Aviv University. He studies machine learning and natural language processing, including the theory of learned representations, attention, and interpretable reasoning.
University of Oxford
Fazl Barez is a senior researcher at the University of Oxford and leads the TSG Lab. His work connects interpretability, evaluation, model control, and machine unlearning with the technical foundations needed for effective AI governance.
EPFL
Robert West is an associate professor at EPFL, where he leads the Data Science and AI Lab. His work spans AI, natural language processing, and computational social science, including safe AI and structured human–AI collaboration.
Cambridge Boston Alignment Initiative
Emre Yavuz directs programs at the Cambridge Boston Alignment Initiative. He builds research communities and talent pathways for AI safety and biosecurity, connecting researchers through fellowships, training, and cross-institutional collaboration.
Institute for Responsible Superintelligence
Adam Tauman Kalai is a cofounder of the Institute for Responsible Superintelligence and a former research scientist at OpenAI. His work spans AI safety and ethics, algorithms, fairness, AI theory, game theory, and crowdsourcing.
Tara Research · Mila
Tommaso Tosato is a cofounder of Tara Research and a postdoctoral fellow at Mila. His research measures honesty in frontier agents and evaluates white-box techniques, including activation steering, for detecting and reducing deceptive behavior.
IBM Research
Erik Miehling is a research scientist in the human-centered AI group at IBM Research. His work focuses on generative-model steerability, multi-agent coordination and alignment, and the dynamics of human–AI interaction.
Meta
Diego Garcia-Olano is a senior research scientist at Meta. His work spans safety alignment, language-model interpretability, and the evaluation and mitigation of risks including hallucination, memorization, and sensitive-data exposure.
Google DeepMind
Been Kim is a director and principal scientist at Google DeepMind. Her research develops interpretability methods that help people understand and learn from complex models, including TCAV, agentic interpretability, and work transferring superhuman chess concepts from AlphaZero to grandmasters.
Columbia University
John Hewitt is an assistant professor of computer science at Columbia University. He studies the safety, control, and efficacy of neural language systems by investigating their internal mechanisms and surprising generalization behaviors.
University of Chicago · Google DeepMind
Victor Veitch is an assistant professor of data science and statistics at the University of Chicago and a part-time research scientist at Google DeepMind. His work develops causal approaches to representation learning and methods for building and evaluating safe, credible AI systems.
Boston University
Najoung Kim is an assistant professor of linguistics and affiliated faculty in computer science at Boston University. She uses computational and experimental methods to study how human and machine learners represent meaning and generalize beyond their experience.
McGill University · Mila · Mistral
Siva Reddy is an associate professor of computer science and linguistics at McGill University, a Canada CIFAR AI Chair at Mila, and an AI scientist at Mistral. His research builds reliable and useful language models and agents across representation learning, memory, reasoning, evaluation, and alignment.
ETH Zurich
Ryan Cotterell is an assistant professor of computer science at ETH Zurich. His research connects natural language processing with formal-language theory, information theory, and linguistics to understand the computational structure and representations learned by language models.
Tilburg University
Afra Alishahi is a professor of computational linguistics at Tilburg University. She develops computational models of language learning and studies grounded language, learning through interaction, and the linguistic representations that emerge inside neural models.
Cadenza Labs
Walter Laurito is a team lead and researcher at Cadenza Labs. His work focuses on white-box model evaluation, interpretability, and AI safety; he previously implemented benchmarks for the UK AI Security Institute through ASET and participated in MATS.
Harvard University · Google
Himabindu Lakkaraju is an assistant professor at Harvard University and a part-time senior staff research scientist at Google. Her research develops methods for AI safety, interpretability, and reliability, with particular emphasis on understanding and evaluating foundation models.
University of Maryland
Sarah Wiegreffe is an assistant professor of computer science at the University of Maryland. Her research develops methods for AI interpretability and explainability, focusing on faithful reasoning traces, monitoring model internals, and making AI systems more transparent, reliable, safe, and useful.
University of North Carolina at Chapel Hill
Mohit Bansal is the John R. and Louise S. Parker Distinguished Professor at UNC–Chapel Hill and director of the MURGe-Lab. His research spans multimodal generative models, reasoning and planning agents, faithful language generation, and interpretable, efficient, and generalizable deep learning.
Neuronpedia · Decode Research
Johnny Lin is the founder of Neuronpedia, an open-source platform for mechanistic interpretability that collaborates with labs, universities, and researchers. He runs Decode Research, the umbrella nonprofit of Neuronpedia that maintains SAELens and other interpretability libraries.
Saarland University · DFKI
Marius Mosbach is an associate professor at Saarland University and a scientific director at the German Research Center for Artificial Intelligence (DFKI). His research studies the capabilities and inner workings of large language models, with a focus on interpretability for model adaptation.
Additional researchers who have joined ICINAi as founding members.