Baihan Lin is a tenure-track professor of AI, psychiatry, and neuroscience at the Icahn School of Medicine at Mount Sinai, where he directs the Bytes of Minds Lab.
He is also a Senior Computer Scientist with the US Department of Veterans Affairs and a Faculty Associate at the Berkman Klein Center. His work sits at the intersection of computational psychiatry and Neuro-AI: building brain- and cognition-inspired models that infer psychiatric and cognitive state from speech, language, and multimodal behavior, and putting them into settings where being wrong has consequences. His group of 20+ researchers, physicians, and engineers works on foundation models for clinical speech and language, interpretable analyses connecting model behavior to cognitive and clinical theory, and systems that run inside real clinical workflows. The lab deployed an LLM voice agent for patient health coaching in a major US hospital network, and co-directs the Therapeutic Alliance AI core of the $20M NIH IMPACT-MH consortium across six sites.
Before Mount Sinai he did machine learning research at Google X, where he was a core inventor on seven US patent filings, and at IBM Research, Amazon, and Microsoft Research. He is the book author of "Privacy & Security in Large Language Models" (O'Reilly, 2025) and "Reinforcement Learning Methods in Speech and Language Technology" (Springer, 2024), with 100+ peer-reviewed publications and patents. He holds a PhD in computational neuroscience from Columbia University and master's degrees in applied mathematics (University of Washington) and data visualization (Parsons School of Design). He is a Bell Labs Prize and XPRIZE finalist. At the Center, his focus is the AI governance question he can speak to from both sides: what happens when models act in psychological settings. Conversational systems are now among the most widely deployed AI in the world and are routinely used for emotional support by people in distress, with almost no agreed standard of evidence for whether they help or harm. He works on making evaluation in these settings construct-valid rather than rubric-matching, on what a deployment-grade safety case would actually require, and on where the line falls between a wellness product and a regulated medical device. He founded and chairs the AI Governance workshop series at AAAI and IJCAI (2024–2026), and welcomes collaboration with BKC colleagues on AI evaluation, oversight, and deployment policy.