VP / Head of Product · Chicago & Remote
Healthcare AI product leader and founder.
Clinical AI rarely stalls on capability. It stalls at the review where someone has to say who owns it when it's wrong. I've taken products through that review at Epic, Doximity, and Transcarent.
- Résumé →
- First product hire at a pre-revenue oncology venture
- Founding product lead of Doximity Dialer
Why me
My work has moved outward one level at a time: whether the system works, whether the product works, whether the business works, whether an institution can rely on it. At Epic that meant fixing failing hospital implementations. At Doximity, making physician calling trustworthy enough for patient care. At Transcarent, building the routing and escalation machinery behind four care programs.
Companies bring me in at exactly that point — increasingly, before it: deciding what to build, which assumption to test first, and what evidence would settle it. The work pays for itself where something already bought is stuck: a pilot that won't convert, a model that can't clear review, a contract nobody can turn on.
Selected work
The identity infrastructure behind trusted telehealth
Built the identity infrastructure that let the office number travel with the physician on their personal phone — dissolving the bind between hospital desk phones and exposed personal numbers. The architecture decision was to make trust a product surface, not a compliance footnote.
Read case study →
Scaling specialty-care product strategy across four clinical programs
Scaled product strategy across four specialty-care programs — Surgery, everyday urgent care, Behavioral Health, and Oncology — by building shared decision infrastructure instead of per-program customization, making routing, ownership, and escalation explicit at every decision point.
Read case study →
What happens after the software is installed
Epic's software was installed and still failing at the accounts I inherited. Owning what happened after go-live meant root-cause analysis, re-implementation, and writing Caché plug-ins to close the gaps the base product left — then building the escalation path so regulatory changes and critical bugs stopped arriving as surprises.
Read case study →Current work
Working tools for the problem above — what an institution has to be able to see before it lets a system act.
- NextConsensus Predicts guideline changes and regulatory decisions before they happen. Each forecast has a deadline and a yes/no outcome, so it gets scored.
- Ambit Keeps a live map of what an AI agent can actually do — the tools, machines, accounts, and permissions it has, and how they fit together.
- Refract Tells you when a source changed and which claims or citations built on it may now be out of date.
- Fast Harm, Slow Repair Measures what happens after an AI system gives a wrong answer — how far the error spreads, how long repair takes, and what stays wrong.
Major product wins
| 35K+ | physicians surveyed | Doximity — I ran the largest national compensation survey to date |
|---|---|---|
| 4 | specialty-care programs launched | Transcarent — Surgery, everyday urgent care, Behavioral Health, Oncology Care |
| 1,200+ | physician users | Andwise — I co-founded it |
| $5M | revenue | Doximity Talent Finder — I grew it 60% YoY |
Organizations I've worked with
Explore
- Building NextConsensus, Ambit, and the work around them.
- Decisions Four calls, with the reasoning as I recorded it before the outcome was known.
- Ethotechnics The method for making consequential AI claims verifiable and correctable.
- Writing Essays on AI governance, institutional power, and cognitive scarcity.
- Build Practice The agent-driven toolchain this site is built and checked with.
- Research The open questions behind the method, and what would settle them.
Email me. For NextConsensus pilots, product leadership conversations, or advisory work.