Research Fellow

This fellowship focuses on an ongoing research project to create, test, and validate AI methods for detecting errors in multimodal AI models, with a current emphasis on mechanistic interpretability techniques in natural language processing. The role involves collaborating with computer scientists and clinicians to design, coordinate, and analyze domain expert annotations and assessments of these methods in clinical medicine. Participants will receive mentorship from clinician and AI faculty at Harvard-affiliated institutions, with opportunities to publish and present at national and international meetings.