Postdoctoral Researcher · McDonough AI Fellow
Dana-Farber Cancer Institute · Harvard Medical School
When should we trust AI with decisions in biomedicine — from patient diagnosis to biological discovery?
My research focuses on building dynamic evaluations grounded in real-world scientific and medical workflows, to benchmark AI readiness for deployment. I also build computational gyms (a.k.a. RL environments) that convert these workflows into training substrates, where AI systems learn through iterative feedback and policy improvement.
This work spans medicine, biology and computer science, and has been published in Nature Medicine, Nature Communications, Cancer Cell and Science Advances, and covered by The Guardian, Science and New Scientist. I also engage with healthcare policy and administration leaders on integrating AI into clinical practice, and write and speak about medical AI for the general public. I am a 2023 Quad Fellow and former Cargill Global Scholar.
Google Scholar, August 2026.
Full list of publications can be found on Google Scholar.
bioRxiv · 2026
AI agents placed inside multi-step discovery workflows across 11 cancer types and 100,000+ patient-derived cells, tested against independent biological evidence.
Nature Medicine · 2025
Medical exam benchmarks such as the USMLE overestimate clinical readiness: under simulated doctor–patient conversation, leading models degrade sharply.
Nature Communications · 2025
BEANIE, a nonparametric method correcting the patient-specific biases that inflate false positives in single-cell cancer data.
Reporting, interviews and expert commentary on medical AI — plus institutional coverage of the underlying research.
Science
AI is starting to beat doctors at making correct diagnoses
The Guardian
“DeepSeek is humane, doctors are more like machines”: my mother’s worrying reliance on AI for health advice
New Scientist
AI chatbots fail to diagnose patients by talking with them
Rest of World
AI chatbots and self-diagnosis in China
Where the research has landed beyond the literature: in deployment policy, and in public understanding of medical AI.
Co-authored a proposal for a transparent labeling system documenting the clinical value trade-offs AI recommendations encode, so health systems, regulators and patients can judge whether a system's priorities match their own.
Engaged with policymakers and scientists across the four Quad nations as part of the inaugural cohort.
Subsequently cited in reports by the U.S. Department of Homeland Security and the Future of Privacy Forum.
Editorial team; wrote for the platform and delivered seminars translating current research for general audiences.
Fellowships, awards and invited scientific talks.
For exceptional postdoctoral candidates pursuing the frontier of AI for metastatic prostate cancer research.
A $50,000 scholarship to pursue graduate studies in STEM for social impact. Selected as one of 100 fellows from more than 3,200 applicants.
Highest GPA in the graduating class of the Department of Biochemical Engineering and Biotechnology.
For exceptional mentorship of incoming first-year students.
A $5,000 award for academic and co-curricular excellence and exceptional leadership. Selected as one of 10 students from India.
Best Paper Award at the symposium, for CRAFT-MD.