BMIF 203

Researchers and clinicians increasingly use AI methods to study biology, generate hypotheses, and support clinical and healthcare decisions. We organize this course around three paradigms in modern AI, namely foundation models, generative models, and AI agents. For each paradigm, we study the underlying architectures and training strategies and examine how design choices affect model performance. Foundation models learn general representations from large datasets, and researchers can adapt them to many downstream tasks. Generative models learn data distributions and generate new samples such as clinical text, images, and molecules. AI agents combine language models with tools, memory, and planning to perform multistep tasks in research and clinical settings. Students apply these models to clinical, imaging, longitudinal, and molecular data and learn to integrate multiple modalities. Evaluation of AI models is a core part of the course, and we also discuss responsible implementation and the requirements for using these systems in research and clinical practice.

BMIF 203 is a course requirement for students in the AIM PhD track. There are limited spots for other students and cross-registration. More details will be shared closer to registration.

View in Course Catalog