01
Real vs. rendered
What runs autonomously today, what still needs a human in the loop, and what is a roadmap promise dressed up as a feature.
Healthcare Private Equity
An operator's read on the AI in your deals.
I'm a physician by training and a product operator who has shipped agentic AI into real clinical workflows. I help healthcare investors tell the AI that works from the AI that only demos, and find the automation upside sitting inside a portfolio company's operations.
Why bring me in
Few people sit where I do: MD, builder, and operator.
MD
Clinical fluencyAn MD who reads workflow
I hold an MD and understand how care actually gets documented, billed, and handed off, so I can judge whether an AI claim survives contact with a real clinic.
↗
In productionAgentic AI, shipped
I've built and deployed agentic AI in a live healthcare setting at Duke through Trase, work described in a public preprint. I evaluate products against what it takes to run in production, not what fits on a slide.
◇
Investor-sideDiligence and roundtables
I've advised investors on healthcare AI through expert consultations and roundtables, so I know the questions that separate a durable advantage from a thin wrapper.
Public reference: co-authored preprint, "Agentic Artificial Intelligence as a Catalyst for Administrative Modernization," medRxiv, 2026.
Where I help
Three points in the deal lifecycle.
Before signing, at the board table, and inside the portfolio company once the check clears.
Diligence
Pressure-test the AI claim
Is the "AI" a real capability or a thin layer on someone else's model? I assess the technology, the data and integration reality, the team, and the roadmap, then translate it into plain risk language for the deal team.
Value creation
Find the automation upside
Across a portfolio the same manual work repeats: intake, documentation, revenue cycle, back office. I map where agents can take real load off headcount and where they can't yet, so the thesis rests on what's actually buildable.
Execution
Help portfolio companies ship
A diligence memo doesn't deploy itself. I work with an operating team to scope, sequence, and stand up the automation the investment case assumed, on a timeline that survives reality.
The lens
How I read an AI claim.
02
Data and integration
Whether the product can actually reach the EHR, the claims system, and the messy data it depends on, or whether that is the unbuilt hard part.
03
Durability
What holds up if a foundation-model vendor ships the same feature next quarter, and what the real switching cost is.