Meng Jiang (Notre Dame): Schmidt AI in Science Speaker Series
Organized by the University of Chicago’s Eric and Wendy Schmidt AI in Science Fellowship Program.
Agenda
3:45pm – 4:00pm: Light refreshments available
4:00pm – 4:45pm: Presentation
4:45pm – 5:00pm: Q&A
Meeting Location
Faraco Family Commons, Room 105
Data Science Institute
5460 S University Ave
Title: Foundation Models for Polymer Material Discovery
Abstract: Polymers underpin technologies ranging from gas separation and thermal insulation to aviation and chemical storage. Yet discovering high-performing polymers remains difficult because the design space is enormous while experimentally labeled data are scarce. This talk presents foundation-model approaches across three central tasks in polymer discovery: virtual screening, inverse design, and retrosynthesis planning. For virtual screening, I will discuss graph neural network methods that combine interpretable rationalization, semi-supervised learning, and imbalance-aware data augmentation, enabling the discovery, synthesis, and experimental validation of novel gas-separation membrane materials. For inverse design, I will introduce Graph Diffusion Transformers for controllable, multi-property molecular generation; DemoDiff, which performs in-context molecular design from molecule-score demonstrations; and CoMole, which uses motif-aware reinforcement learning to make chemically meaningful decisions and align molecular generation with verifiable property rewards. Finally, I will present Llamole, a multimodal large language model that uses graph generative models as tools to jointly design molecules and plan their synthesis routes. These projects illustrate how graph representation models, graph generative models, and multimodal language models can complement one another in data-limited scientific discovery. They also raise a broader question: Should these model classes ultimately be unified into a single scientific foundation model, or do their distinct capabilities remain essential?
Bio: Meng Jiang is the Frank M. Freimann Collegiate Associate Professor of Computer Science and Engineering at the University of Notre Dame. He co-directs the Foundation Models and Applications Lab. He also serves as Program Chair of the Notre Dame-IBM Tech Ethics Lab. His research develops strong, controllable, and multimodal AI, with applications in scientific discovery, engineering, education, and mental health. His work spans foundation models, graph learning, natural language processing, data mining, with an emphasis on molecular and polymer discovery for seven years. He received his B.E. and Ph.D. in Computer Science from Tsinghua University and is a recipient of the NSF CAREER Award.
Seppe Kuehn (UChicago): Schmidt AI in Science Speaker Series
Ashesh Chattopadhyay (UC Santa Cruz): AICE Speaker Series