Introduction

I am an Assistant Professor at Harvard University in the Department of Biomedical Informatics.

My work focuses on pushing the boundaries of AI to revolutionize medicine by developing systems that can interpret medical data, reason through complex problems, and communicate at an expert level. The goal is to create AI doctors that can work independently or alongside human physicians, transforming care delivery worldwide and ensuring everyone has access to high-quality, timely treatment.

I have published more than 100 academic articles garnering more than 25,000 citations, including in Nature, NEJM, and Nature Medicine. My work has been featured in media outlets like NPR, The Washington Post, and Wired. I have been recognized as an awardee of Forbes 30 Under 30 in science ‘22, Innovator Under 35 by MIT Tech Review ‘23, Nature Medicine Early-career Researcher To Watch ‘22, and Google Research Scholar Program. Prior to starting as faculty at Harvard in 2021, I received my B.S., M.S., and Ph.D. degrees, all in Computer Science from Stanford University

Contributions

Expert-Level Medical AI: I have developed medical AI algorithms that diagnose diseases from images at the same level as medical experts. My teams have made breakthroughs in self-supervision and pre-training, creating label-efficient algorithms for various medical imaging modalities. In 2019, an algorithm I developed that detects abnormal heart rhythms from ambulatory electrocardiograms at the level of cardiologists received FDA clearance. CheXzero, a recent milestone, detects diseases from chest X-rays using natural-language descriptions from clinical reports, eliminating the need for manual annotations.

Data Curation and Human-AI Collaboration: I have led the curation of widely used datasets, such as CheXpert for imaging and SQuAD for natural language processing, driving progress in deep learning and highlighting generalization gaps in AI models. My team develops open benchmarks to measure algorithm generalizability across patient populations and imaging modalities. I have also investigated the impact of medical AI on clinicians' performance, optimizing human-AI collaboration in clinical workflows.

Generalist Medical AI Systems: My group pioneers the development of Generalist Medical AI systems that resemble doctors in their ability to reason through diverse medical tasks, incorporate multiple data modalities, and communicate in natural language. We have developed methods to combine data sources like images, sensors, and language to improve decision-making and generalization. We also build generative AI models that interpret medical images in natural language and interact with clinicians, including a copilot approach to radiology report generation, where AI models provide initial drafts for clinicians to edit, streamlining the diagnostic process.

Community Efforts

I am the program director of the Medical AI Bootcamp, a program for closely mentored research at the intersection of AI and Medicine that is open to students at Harvard & Stanford, and to medical doctors around the world. I proposed, designed, and instructed CS197: AI Research Experiences at Harvard, a course in applied deep learning research. Beyond Harvard, I previously designed and instructed a massive open online course (MOOC) series called Artificial Intelligence for Medicine (AI4M) on Coursera, which has over 60K students and enables people with diverse backgrounds to enter the medical AI space. I co-founded the Doctor Penguin newsletter, a weekly newsletter summarizing major new developments in medical AI research for over 5K readers. I also co-founded the AI Health Podcast, which has had over 40K listener episode downloads, and highlights industry and scientific developments at the intersection of AI and health.