Speaker: Dr. Adnan Jafar

Topic: Towards Human-Centric AI-Agents for Personalized Healthcare

Abstract: Current cancer treatment planning relies on a trial-and-error process, where human planners use specialized medical software to generate treatment plans for physician approval. The quality of these plans heavily depends on the planner’s expertise and is often constrained by high patient volumes and limited time. This process is further slowed by repeated planner–physician communication during iterative plan refinement, potentially delaying treatment and affecting outcomes. My research addresses these challenges by developing human-centric intelligent agents that use large language models and multi-modal data to automate the treatment planning workflow under physician oversight.
In parallel, I also work on diabetes care, particularly for individuals with Type 1 diabetes—a chronic condition requiring lifelong insulin therapy. I have developed reinforcement learning algorithms to personalize insulin dosing for high-fat meals and postprandial aerobic exercise. These algorithms were clinically validated in a 16-week study, demonstrating significant improvements in postprandial glucose levels compared to baseline.

Bio: Dr. Adnan Jafar is a Postdoctoral Research Scholar in the Department of Radiation Oncology at Johns Hopkins University. As part of his postdoctoral research, he is working on developing advanced large language model–based systems for automated and personalized cancer treatment planning. His active research interests span both cancer and diabetes, with a long-term goal of advancing personalized care in both domains. He earned his Ph.D. in Biomedical Engineering from McGill University, where he developed and clinically tested novel reinforcement learning algorithms for personalized diabetes management. Prior to that, he was a researcher at the University of Sannio Italy, working on the modeling and control of cyber-physical systems. Dr. Jafar brings one year of industry experience and two years of university-level teaching experience. He also holds a patent on an AI-based diabetes product, which was acquired by a biomedical company.