HARVARD MEDICAL SCHOOL · DEPARTMENT OF BIOMEDICAL INFORMATICS

Medical AI for complex clinical work.

We study how AI can reason from medical evidence, make decisions over time, and assist with physical clinical procedures.

Explore three research programs

THE PUBLICATION RECORD

GOOGLE SCHOLAR · UPDATED AUG 9, 2026

publications
177
citations
18,618
since the lab opened
2021–2026

PUBLISHED INNature · Nature Medicine · Nature Biotechnology · Cell · The Lancet · NEJM AI · JACC · Radiology

Browse publications Google Scholar

CURRENT RESEARCH

Three programs.

Each starts with a different kind of clinical work—and a different standard that today's systems do not yet meet.

  1. 01

    Grounded reasoning and evaluation

    Can a model show what evidence supports its answer—and can we measure when it is wrong?

    1. Locate
    2. Reason
    3. Measure
    4. Stress-test

  2. 02

    Clinical agents in simulated care

    What happens when an AI must care for patients over time, rather than answer one isolated question?

    1. Observe
    2. Act
    3. Environment changes
    4. Reassess

  3. 03

    Procedural learning and robotic assistance

    Can AI learn the spatial and physical structure of a clinical procedure well enough to assist?

    1. Perceive
    2. Capture
    3. Assist

RECENT FROM THE LAB

One project from each program.

  1. Figure from the paper: grounding free-text CT findings to pixel-level segmentations
    01Grounding & evaluation

    ReXGroundingCT

    Connects free-text radiology findings to expert-verified, pixel-level evidence in three-dimensional chest CT.

    NEJM AI · 2026

  2. Figure from the paper: changing patient and hospital signals over time
    02Clinical agents

    Clinical Environment Simulator

    Proposes a persistent digital hospital where agent decisions alter patients, resources, and subsequent choices.

    Nature Medicine · 2026

  3. Figure from the paper: imitation-learning robot policies evaluated on suture following
    03Procedures & robotics

    Robot assistance in open surgery

    Evaluates general-purpose imitation-learning policies on a bounded collaborative task performed alongside a surgeon.

    Preprint · 2026

PEOPLE

A compact, interdisciplinary lab.

Computer scientists and clinicians working across three research programs.

Faculty

Postdocs

PhD students

Master's students

Affiliates

View full profiles and alumni

JOIN THE LAB

Different routes, clearly separated.

  1. 01Postdoctoral and predoctoral researchersApply directly to the lab through a structured application.
  2. 02Prospective PhD studentsApply through the Harvard AI in Medicine PhD program.
  3. 03Medical AI BootcampA separate mentored research program with its own application process.