Towards AI Scientists: Scientific Agents and Biomedical World Models
Zhenglun Kong · Computer Science Colloquium Series
| Speaker | Zhenglun Kong |
| Date | Friday, October 23, 2:15 – 3:05 pm |
| Location | 307 Love |
Abstract
Recent advances in foundation models and AI agents are creating new opportunities to accelerate scientific discovery. However, moving beyond general-purpose language models requires AI systems that can effectively use scientific tools, integrate diverse sources of knowledge, and reason about complex biological processes.
In this talk, Zhenglun will introduce how AI agents can integrate scientific tools, biomedical knowledge, and multi-step reasoning to tackle complex research tasks. He will highlight recent efforts to build scalable and extensible ecosystems that enable AI agents to access and coordinate thousands of scientific tools, supporting tasks ranging from knowledge retrieval and data analysis to therapeutic reasoning and research planning.
Zhenglun will also discuss the emerging direction of biomedical world models, which aim to move beyond reasoning over existing knowledge toward predicting how biological systems respond to interventions and evolve over time. He will explore the opportunities and challenges of developing models that can simulate biological dynamics, predict counterfactual outcomes, and guide experimental design.
About the speaker
Zhenglun Kong is a tenure-track Assistant Professor of Electrical and Computer Engineering at the University of Miami. Previously, he was a Research Fellow at Harvard University, where he worked on foundation models and AI agents for scientific discovery. He received his Ph.D. and M.S. degrees from Northeastern University and his bachelor’s degree from Huazhong University of Science and Technology. He has also held research internships at Microsoft Research, Arm, and Samsung Research, and was named a Machine Learning and Systems Rising Star in 2024. His research focuses on efficient and autonomous AI, spanning foundation models, AI agents, and AI for scientific discovery. His recent work includes developing scientific agents for biomedical reasoning and multimodal models for biological systems. He also works on scalable and resource-efficient AI, including post-training and inference optimization, model compression, and memory-efficient architectures.