AI Center Seminar - AI for Science series - Dr. Alex Tong

The talk is organized by the EPFL AI Center and the EPFL Laboratory of Protein Design and Immunoengineering (LPDI) as part of the AI for Science seminar series.

Hosting professor: Prof. Bruno Correia (LPDI)

Title
Controlling Generative Models

Abstract
Generative modeling excels at mimicking data distributions, but scientific discovery often requires sampling from a modified distribution. In this talk, I will motivate and present a framework for controlling flow models to control sampling on a distribution level. I will demonstrate how this approach can be adapted to two distinct challenges in the natural sciences: (1) guiding generation to control for multiple functional properties for design, and (2) steering dynamics to sample from unnormalized energy densities for equilibrium sampling. Together, these methods illustrate the versatility of flows for both finding new, optimized structures and exploring the physical landscapes of existing ones.

Bio
Alexander Tong is a Principal Investigator at Aithyra in Vienna, Austria. His research sits at the intersection of machine learning and biology, with a focus on generative modeling, flow matching, and optimal transport. He applies these techniques to problems in protein design and cellular dynamics. Previously, Alex was a postdoctoral researcher at Mila in Montreal, where he worked with Yoshua Bengio, and he completed his PhD in Computer Science at Yale University in 2021. He also co-founded Dreamfold, a company focused on generative protein design.