Joo Heung Yoon, MD

I have been interested in identifying hidden patterns of pathologic state with machine learning, and developing prediction models with using a large-scale clinical data. With the collaborations with the School of Machine Learning, Carnegie Mellon University, I have been building various prediction models using supervised ML as well as deep neural network. My exploration on the multigranular, multimodal clinical data allowed our research team to identify risks for impending shock and hypovolemia, and prompted to generate reinforcement learning-driven simulation of shock management. My current scope of work expands to the development of physiologic foundation model and leveraging few state-of-the-art large language models for unconventional clinical data utilization.

As a critical care physician, I spend most of my time in the Intensive Care Unit (ICU). My clinical interests are shock and resuscitation, especially on identifying etiologies of physiologic deterioration from various acute illnesses.

In the ICU, I provide structural teaching to house staffs and medical students at a daily basis. I like to teach cardiopulmonary physiology and resuscitation, that includes pulmonary mechanics, use of vasopressors/inotropes, response to resuscitation, and post-cardiac arrest care. Also I teach clinical informatics to medical students and bedside rounding teams as needed.