This research focuses on the comprehensive analysis of diverse multimodal biomedical data—such as endoscopic images, biosignals, and electronic health records (EHR)—to gain deeper insights into disease mechanisms and advance personalized medicine. Our goal is to leverage hospital real-world data (RWD) to broaden the clinical applications of Large Multimodal Models (LMMs) and agentic AI, ultimately validating them in real-world clinical workflows.

This research focuses on embodying AI in the physical world of clinical practice, enabling systems such as robotic arms and humanoid robots to autonomously perform medical procedures. By integrating multimodal perception, physics-aware modeling, and real-time control, we aim to develop embodied agents that can safely interact with patients and clinical environments. Our goal is to bridge the gap between AI and physical execution, ultimately delivering autonomous systems that operate reliably alongside clinicians in real-world clinical workflows.