Kyungsu Kim
Assistant Professor Seoul National University
Kyungsu Kim, PhD, is an Assistant Professor specializing in Biomedical and Core AI at Seoul National University, with appointments in the School of Transdisciplinary Innovations, the Department of Biomedical Sciences within the College of Medicine, and the Interdisciplinary Program in Artificial Intelligence within the College of Engineering. His research focuses on developing foundational generative AI methodologies—including multimodal diffusion models and stochastic interpolants—and applying them to biomedical science, structural biology, multimodal and agentic systems, and drug discovery. Previously, he held research positions at Harvard Medical School and Massachusetts General Hospital, Samsung Medical Center, and KAIST. Dr. Kim has published extensively at leading AI conferences and in prominent biomedical journals and has served as an Area Chair for premier AI and machine learning conferences, including NeurIPS, ICLR, and ICML.
Seminars
AI has truly become the buzzword as the tool to improve the quality of, and accelerate, drug discovery timelines. But how impactful are AI/ML discovery tools, and what has been proven experimentally so far to demonstrate reliability?
Join this workshop to break down the applications and bottlenecks of AI/ML for de novo discovery, and learn how best to blend in silico and wet lab expertise.
Workshop Highlights Include:
- Seeing through the AI hype: Breaking down different augmentations of AI/ML tools to improve the quality and reduce cost and timelines of drug discovery workflows
- Understanding what has been shown to be effective so far through experimental validation, and what areas remain challenging – what are the core hesitancies to adopt AI/ML discovery tools?
- Limits and and failures of structure prediction tools: Understanding methods to steer AI/ML models to generate transition-state conformations without further training
- Integrating wet lab and computational insights to unlock mutual benefit towards best-in-class drugs
- Contextualizing advances in AI for structure-based drug design from proteincomplex structure prediction to multimodal and 3D diffusion-based molecular generation, while examining the remaining limitations of monolithic models
- Leveraging state-of-the-art multimodal and compositional diffusion AI to overcome bottlenecks in in silico structure-based drug design and advance ligand-binding modeling, scaffold decoration, and lead optimization across modalities
- Weighing up strengths and limitations of AI methodology for de novo drug discovery and highlighting the need for future experimental validation
