Adam Oberman

Adam Oberman is a Full Professor at McGill University's Department of Mathematics and Statistics, affiliated with the Faculty of Science. He holds the Canada CIFAR AI Chair and is an associate member of Mila, Quebec's AI institute. His academic career includes roles at Simon Fraser University (2004-2012) and postdoctoral research at the University of Texas at Austin. His research spans applied mathematics, focusing on numerical methods for PDEs, optimal transportation, and machine learning. Key areas include adversarial robustness, generative models (e.g., normalizing flows), and AI safety. He has contributed to foundational work on PDE solvers, regularization techniques, and neural network optimization. Notable collaborations include projects on AI safety with Yoshua Bengio and work on ethical AI frameworks.

Education: PhD from the University of Chicago, postdoctoral training at UT Austin. Research highlights include the development of EuclidNets for efficient neural network inference, contributions to multi-resolution normalizing flows, and theoretical advancements in reinforcement learning representation generalization. His work bridges mathematical rigor with practical machine learning challenges, addressing issues like adversarial attacks and algorithmic fairness. Teaching includes advanced courses on machine learning theory and numerical PDEs.

Key contributions extend to numerical methods for nonlinear PDEs, including Monge-Ampère equations and optimal transport problems. He has pioneered techniques like filtered schemes for Hamilton-Jacobi equations and regularization approaches for deep learning optimization. Recent efforts focus on mitigating catastrophic risks in AI through frameworks like Scientist AI and collaborations with Safe AI for Humanity (SAIFH).