AI-driven Scientific Discovery

We study the computational foundations of design. Across disciplines, designing complex physical systems, from molecules and materials to biological systems and engineered devices, requires solving similar fundamental computational challenges: how to represent vast design spaces, how to evaluate the behavior and function of candidate designs, and how to efficiently search for solutions that satisfy multiple objectives and constraints. Our research develops general computational frameworks that integrate machine learning, physical simulation, optimization, and scientific experimentation to address these challenges. By advancing representation learning, surrogate modeling, differentiable simulation, Bayesian and evolutionary optimization, and autonomous reasoning, we create algorithms that transform design from a process of trial and error into one of prediction, computation, and increasingly autonomous discovery.

News