Watch the recording here

Zoom.

Date: March 6

Time: 2 p.m.

Location: LINC 268

Abstract: The rapid growth of multimodal biomedical data — spanning imaging, genomics, clinical text, and molecular structures — presents unprecedented opportunities for advancing scientific discovery. However, effectively integrating these heterogeneous data sources remains a fundamental challenge. In this talk, I will present our recent work addressing novel challenges in multimodal biomedical learning. First, I will introduce our framework for flexible and adaptive multimodal learning, designed to operate with any arbitrary combination of modalities. Unlike conventional approaches that assume fixed modality availability, our method gracefully handles missing, incomplete, or dynamically varying inputs, making it well-suited for real-world clinical and research settings where data availability is often unpredictable. Second, I will discuss our efforts toward responsible multimodal learning guided by biological priors, ensuring that model behavior aligns with established domain knowledge and yields interpretable, trustworthy predictions. By embedding biological constraints into the learning process, we can improve generalization, reduce spurious correlations, and promote scientific rigor. Together, these contributions point toward a future where multimodal AI systems can reliably and responsibly accelerate biomedical discoveries.

Bio: Tianlong Chen is an assistant professor in the Department of Computer Science at the University of North Carolina at Chapel Hill and is the Chief AI Scientist at hireEZ. He received his Ph.D. degree from ECE at UT Austin in 2023 and had a postdoc at MIT and Harvard in 2024. His research focuses on efficient and reliable machine learning, large language model agents, multimodal learning, and AI for BioScience. He received the Meta Research Award, Amazon Research Award, Cisco Faculty Award, IBM-UNC Junior Faculty Development Award, and several best/outstanding paper awards from various conferences and workshops like SIGBio, NAACL, LoG, AMIA, ACM MM, AAAI, NeurIPS, etc. He regularly serves as conference (senior) area chairs, journal editors, invited speakers, tutorial/workshop organizers, various panelist positions, and reviewers. His research is gratefully supported by NIH, NSF, DOE, as well as dozens of industry and university grants.