Healthcare product & systems leader
Healthcare product and systems leader building accountability infrastructure for high-stakes AI.
Fourteen years shipping products where a wrong automated decision is measured in patient outcomes — Epic, Doximity, Transcarent, Andwise. Now building the evaluation, review, escalation, and correction systems that let institutions safely rely on AI.
Current focus VP Product — healthcare AI — Also open to selective founding product roles in healthcare AI.
How this work accumulated
In retrospect, one operational problem kept recurring: the higher the stakes, the harder it got to trace, question, or undo a decision.
-
Research assistant modeling RNA folding dynamics, sRNA-mRNA interactions, and Hfq-mediated gene regulation. Co-authored a book chapter in the ACS Symposium Series at 19.
Modeling a free-energy barrier is a search for the step that decides whether a process completes or stalls. The binding constraint is rarely the part of the system that draws attention.
-
Configured EHR workflows, decision-support alerting, and quality-measurement rules across hospital go-lives.
A single misrouted alert can bury a critical lab result. The workflow around the clinician was almost always the constraint — not the clinician.
-
Early product lead on Dialer — verified office caller ID, consent copy, fallback paths for missed connections, and audit-ready call logs.
Patients answered once the call could prove who it came from. Trust is a product decision: what a system claims about itself determines whether anyone acts on it.
-
Directed digital products and strategy for an oncology information and navigation platform serving 30MM annual visitors.
Translating clinical protocol into steps a frightened family can follow, at scale, means the guidance itself becomes the risk surface. Guidance is only as good as whoever owns the next step.
-
Led product across four specialty-care programs — a unified member record, clinician-led routing rules, escalation paths, and care-plan completion dashboards.
Routing output without a workflow owner is just a suggestion. Making ownership visible cut the exception queue.
-
Co-founder and CEO. Built review routing, escalation clocks, accountable sign-off, and a 50+ physician medical advisory board for physician-facing financial guidance.
Sensitive recommendations need accountable sign-off and inspectable provenance — not a confidence score.
-
Frozen-evidence forecasting for guideline, regulatory, and coverage actions; a deterministic evidence-traceability engine; and a published operating model for authorization, challenge, and correction.
The higher the stakes, the harder it becomes to trace, question, or undo a decision. This work makes traceability, challenge, and correction explicit infrastructure instead of institutional goodwill.
Case studies
Product methodology: Ethotechnics
The operating method behind the current work: decision rights, review boundaries, escalation paths, and correction mechanisms for AI deployment. Read the framework →
Organizations I've worked with
Flagship project
Current work
One flagship, and the infrastructure, method, and analysis practice it runs on.
Powers NextConsensus — verification infrastructure
Refract
Open-source (CC0-1.0) engine that checks whether a claim still holds against its sources, producing structured change events with full provenance records.
Governs NextConsensus — operating method
Ethotechnics
The operating method behind the flagship work: decision rights, review boundaries, escalation paths, and correction mechanisms.
Writing and analysis
The Crumple Zone
Essays on eliminating operational friction and bridging the gap between automated recommendations and clinical execution.
Labs — prototype
Capability Graph
Tracking what an AI agent can do, what may be decaying, and the dependencies behind new capabilities.
I'm exploring VP Product leadership in healthcare AI and clinical decision infrastructure. If you're hiring for one — or weighing how an AI system will hold up under clinical deployment and institutional review — let's talk.