Speakers
Invited Speakers
We are happy to feature the following invited speakers (listed alphabetically by last name):
Gintare Karolina Dziugaite
Affiliation: Google DeepMind, Mila, McGill University
Biography: Gintare Karolina Dziugaite is a senior research scientist at Google DeepMind, based in Toronto, an adjunct professor in the McGill University School of Computer Science, and an associate industry member of Mila, the Quebec AI Institute. Prior to joining Google, she led the Trustworthy AI program at Element AI/ServiceNow. Her research combines theory and empiricism on deep learning generalization and compression, including PAC-Bayes perspectives that connect weight-space properties to generalization. She has co-authored many papers on understanding linear mode connectivity, model merging, and symmetries.
Stefanie Jegelka
Affiliation: MIT and TU Munich
Biography: Stefanie Jegelka is an Associate Professor of Electrical Engineering and Computer Science at MIT and a member of CSAIL. Her research develops principled machine learning methods grounded in optimization and discrete mathematics, with applications spanning structured prediction, robustness, and large-scale decision-making. Her recent works focus on the impact of neural parameter symmetries and the development of equivariant models for weight-space learning. Her work has received multiple best paper awards and honors, including the NSF CAREER Award and Sloan Research Fellowship.
Zechun Liu
Biography: Zechun Liu is a Staff Research Scientist and Tech Lead at Meta. Her research focuses on improving the efficiency and deployability of foundation models through architectural optimization, low-bit quantization, and sparsity. Specifically, she is interested in using deep learning to solve practical industry problems such as resource constraints and the trade-off between compute and accuracy. Her recent work focused on leveraging weight-space symmetries for quantization (SpinQuant), a state-of-the-art method for LLM quantization.
Sidak Pal Singh
Biography: Sidak Pal Singh is a Research Scientist at Google DeepMind whose research investigates weight-space symmetries and their role in model merging. His research develops principled alignment methods based on optimal transport and permutation symmetries, enabling effective merging by accounting for symmetry-induced redundancies in parameter space. His recent work extends linear mode connectivity to transformers, demonstrating how symmetry-aware alignment enables merging and interpolation of large-scale foundation models.
Robin Walters
Affiliation: Northeastern University
Biography: Robin Walters is an Assistant Professor of Computer Science at Northeastern University and the director of the Geometric Learning Lab. His research focuses on symmetries in deep learning, spanning theory through the lens of symmetry, symmetry-aware optimization methods, the development of new equivariant models, and applications of equivariant neural networks across science and engineering. Walters is a visiting fellow at the Boston Dynamics AI Institute, where he develops equivariant neural networks for sample efficient robot perception and manipulation.