Events

CILVR Seminar: Optimizing AI agentic systems for medicine

Speaker: Sheng Liu

Location: 60 Fifth Avenue, Room 7th floor open space
Videoconference link: https://nyu.zoom.us/s/92194531429

Date: Wednesday, April 16, 2025

This presentation explores the transformative potential of artificial intelligence (AI) in medicine through two cutting-edge advancements. First, we introduce GPT-RadPlan, an AI-powered radiotherapy treatment planning system that fully automates the planning process. By leveraging large multimodal language models, GPT-RadPlan consistently outperforms expert physicians and dosimetrists, setting a new standard for clinical accuracy and efficiency. Despite its success, optimizing such compound agentic systems remains a significant challenge. To address this, the second part of the talk presents TextGrad, a novel framework for automatically optimizing agentic AI systems using natural language feedback from large language models. TextGrad is broadly applicable across domains, including reasoning, and biomedical research, enabling more interpretable and controllable AI behavior. Together, these innovations highlight the growing role of AI in enhancing clinical decision-making and accelerating biomedical discovery.