Yilun Zhang
Physician turned AI builder. I work on healthcare's hardest problems with frontier AI. π©Ίπ€β‘
currently
building and exploring the intersection of applying frontier AI to solving some of healthcare's stickiest and toughest problems.
I'm writing about that journey as I go β building and learning in public. Stay tuned...
the road so far
Two tracks, run in parallel for over a decade — medicine and building — until they merged. Cards span the years they actually covered.
2025 β 2026
Atropos Health Β· AI Product Manager, R&D
The tracks converge: Worked on greenfield projects within R&D to deploy LLM evaluation systems across all AI verticals and developed agentic systems for automated evaluation using frontier methods.
2025 medicine
Mount Sinai Health System Β· EM Resident
Emergency medicine at Mt Sinai West/Morningside β my last clinical stop before going all-in on AI.
2023 β 2025 building
Microsoft Γ Mass General Hospital Β· Software Engineer / Researcher
Joint ACR/RSNA Common Data Element project with Dr. Tarik Alkasab: built an open-source framework for running local, on-premise LLMs over radiology report corpora to grow a standardized ontology of imaging findings β distributed, humans in the loop. FastAPI, MongoDB, LanceDB, Pydantic.
2024 β 2025 medicine
Rutgers / RWJ Barnabas Health Β· Resident Physician
Vascular surgery, EM, and critical care β while building the ED's AI infrastructure on local LLMs: automated residency-application review and a medication-reconciliation tool aimed at a seven-figure annual problem.
2017 β 2024 medicine
Tufts University School of Medicine Β· M.D.
Along the way, built radiology tooling at Tufts MC and Lahey: one-press report signing outside PowerScribe/PACS, and a RAG-based "Radiology Helper Bot" trained on internal protocols.
2020 β 2023 building
Surgical Informatics Lab, HMS Β· Research Fellow
Computer vision + NLP on open surgical video: curated the AVOS dataset β 2,000 surgical videos from 200,000 YouTube candidates β and automated evaluation of surgical resident technique. Published in JAMA Surgery.
Meanwhile (2021β22): founded InfUr at BU's medical accelerator, helping patients pick where to seek care with real-time utilization and CMS outcome data. Led a team of three to a top-3 finish.
2018 β 2020 building
SurgiBox Β· Market Strategy & BD
Global market research, regulatory landscaping, and board-level reporting for a med-device startup making surgery safe anywhere.
2016 β 2017 building
Albert Einstein College of Medicine Β· Research Technician
EEG studies of autism and Rett syndrome in a medically underserved community.
2015 β 2016 medicine
Minerva Fellowship Β· Jinan, China
A year teaching music therapy to autistic children, and studying how culture shapes perceptions of autism.
2011 β 2015 medicine
Union College Β· BS Neuroscience, summa cum laude
Varsity track & field. First research: an honors thesis on binary metrical processing in auditory neuroscience.
2009 β 2013 building
New York Medical College Β· Research Assistant
Wet-lab pharmacology: bladder cancer cell signaling and 20-HETE pathways.
writing
Trying a new thing to write about my journey as I build and learn in public. Read the blog β
-
2026-06-09
Hello, world!
Initializing...
selected talks & presentations
- Developing a Standardized Ontology for Imaging Findings Using Large Language Models with Crowd-Sourced Iteration and Validation
Society for Imaging Informatics in Medicine - Using Computer Vision and Natural Language Processing Techniques to Curate and Characterize Open Surgical Videos on YouTube
American College of Surgeons - Development of a Tool for Automated Classification of Surgical Videos for the Curation of an Educational Surgical Video Repository
Association for Surgical Education - Training Non-Surgical Experts to Annotate Open-Source Surgical Videos for Machine Learning
Academic Surgical Congress - Leveraging a Real-Time Spatiotemporal AI Model for Surgical Resident Training During the COVID-19 Pandemic
13th Annual Silverman Symposium - A Not-So-"Frozen Shoulder": A Case Report of a Fractured Central Venous Catheter with an Unusual Presentation
American Academy of Emergency Medicine
publications
- Analyzing surgical technique in diverse open surgical videos with multi-task machine learning
JAMA Surgery, 2023 - State of the Heart: A pictorial review of classic and novel cardiac devices
RSNA Educational Exhibit, 2023 - Scaling Up Efforts for a Notable Demographic: Diagnostic Challenges and Practical Considerations for Imaging of Patients with Obesity
RSNA Educational Exhibit, 2023 - The Majority of Patients Aged 40 and Older Having Allograft Anterior Cruciate Ligament Reconstruction Achieve a Patient Acceptable Symptomatic State
Arthroscopy, 2022 - A real-time spatiotemporal AI model analyzes skill in open surgical videos
arXiv:2112.07219, 2021 - Positive Effects of Heme Oxygenase Upregulation on Adiposity and Vascular Dysfunction: Gene Targeting vs. Pharmacologic Therapy
Int. J. of Molecular Sciences, 2019 - Keeping a Pulse on Autism Spectrum Disorder in the East: A Case Study of ASD Awareness and "Face" in China
Assoc. of Medical Humanities Int. Journal, 2016 - Antioxidants Condition Pleiotropic Vascular Responses to Exogenous H2O2: Role of Modulation of Vascular TP Receptors and the Heme Oxygenase System
Antioxidants & Redox Signaling, 2013 - Androgen-dependent hypertension is mediated by 20-HETE-induced vascular dysfunction: role of IΞΊB Kinase
Hypertension, 2011 - NO and 20-HETE of Vascular Origin are Reciprocal Modulators of the Co-Mediated Dilator Action of Exogenous H2O2 in Renal Arteries
Hypertension, 2010
Full list on Google Scholar.
misc
- For eight years I coached ice hockey for children and adults with developmental disabilities with the New York Raptors.
- I speak fluent Mandarin and conversational Spanish.
- Off the clock: motorsports, ice hockey, and web development (this site counts).
- 0 frameworks were used to build this site β pure HTML and CSS.