// SELECTED WORK
The constant was never the domain, it was the method. Each case is framed the same way: the bet, the call I made, what happened, and what it proves. Click any case study to expand the full narrative.
Every telehealth competitor was solving for video quality, hardware integration, and clinical feature breadth, the problems a procurement committee grades on. After shadowing workflows across five service lines and interviewing more than 30 clinicians, I found the real bottleneck lived in the 30 seconds before a call even started: friction at consult initiation, not anything a feature comparison would surface.
The team wanted to hire four more engineers and expand to three or four service lines at once, chasing demand from fourteen service lines asking in. I made the opposite call: pause new feature work for two months, borrow a designer and a product fellow instead of adding headcount, and spend six months validating the workflow on just two service lines before building anything horizontal. I also deferred a mobile app, deeper EMR integration, and an AI layer, all real requests, until the platform foundation held.
Response time held near real-time across eight external partner contract renewals. The platform scaled to 40,000+ consults across six service lines and 30+ locations in a 40-plus hospital system. I also spent a year chasing a standalone venture raise; investors were right that we had no proof of demand outside our home system, but wrong to lump us in with the telehealth incumbents then collapsing. Separating those two conclusions is what reset the strategy from raising capital to finding a partner, and from there to acquisition by a strategic acquirer.
The standard of care wasn't just outdated, it was compounding against itself: paper assessments required annual provider re-certification, the reimbursement paperwork was so heavy that most qualifying visits never got billed, and providers avoided diagnosing a disease with no cure. A digital version of the same test already existed and solved none of it, it digitized the form, not the workflow. I built evaluation criteria before looking at a single vendor, specifically to avoid rationalizing a favorite, then screened 22 companies globally. Almost all of them clustered after diagnosis; the gap was integration before it.
Building in-house was the easy option to reject: the real moat wasn't the software, it was validated science, FDA clearance, and adoption at 100+ primary-care practices with no health-system backing, proof the hardest problem, getting providers to change behavior, was already solved. So I invested and integrated instead of building. Adoption nearly died forty minutes into a pivotal meeting when an influential clinical stakeholder miscategorized the tool as a specialist-only assessment. Rather than argue the clinical case, I addressed it directly with a number: the reimbursement the health system was already leaving on the table. I advised the board as an observer through the rest of the scale-up.
That reframe held. The stakeholder's conditional approval required a formal usability study run through the institution's own quality-improvement process, not an informal pilot, so skeptics couldn't dismiss the evidence later. It passed, and the tool scaled to 50+ Family Medicine sites and 200+ providers, integrated into the market-leading EHR, with a patient-stratification protocol connecting screened patients to specialist care. What unlocked it wasn't new clinical evidence, it was naming money the health system already qualified for and wasn't collecting, existing reimbursement recovered by fixing the workflow, not new revenue. Four years later, the same stakeholder who nearly killed it asked me to expand it into the specialist workflows he'd originally been protecting.
The gap wasn't a shortage of chronic-pain products, it was that a proven model was trapped. My home health system had already spent twelve years and 40,000 patients validating a physician-supervised, non-physician-led care model in eleven clinics, cutting ED visits by roughly 15%, but it only existed in person, with a six-to-twelve week wait. Separately, primary-care research (35 interviews, nearly 1,000 survey responses) surfaced the real job to be done: physicians didn't want a better tool, they wanted the condition off their plate entirely. That reframed the venture from "give doctors better tools" to "take the patient over."
Clinical leadership wanted a pure software layer; a fellow venture builder wanted to avoid a services-heavy company entirely. I made a third call: build a tech-enabled services company, justified by a model that scaled from about 40% margin on day one to 60% as automation matured, using services to earn the trust and outcomes data that would later fund a defensible technology moat. When the MVP, billed under the referring physician's own tax ID, met unanimous frontline resistance, I didn't push change management harder. I diagnosed it as structural: billing under the physician's ID kept the exact liability we were supposed to remove. The fix was to bill under our own tax ID as an accountable entity, so the physician's job shrank to "just refer."
That single pivot let one signature open a practice network instead of selling practice by practice, on the same timeline as the failed MVP. Three more health systems signed within three to four months of launch. A joint research collaboration with my home health system showed 50% fewer ED visits and 67% fewer hospitalizations in one cohort, and 59% fewer hospitalizations and 31% fewer urgent-care visits in another, across a 40,000-patient, twelve-to-eighteen-month dataset. Financing is milestone-gated, not calendar-based: the next tranche requires five signed contracts, $2M in ARR, and a working predictive model.
// EXPERIENCE
Built and scaled enterprise health platforms, originated venture theses across priority clinical domains, advised portfolio companies as a board observer, and led the commercialization and exit of an internal platform to a strategic acquirer.
Prototyped a developer-insights platform giving engineering-productivity visibility across ~3,000 engineers.
Built a human-in-the-loop data-validation platform; re-architected the enterprise data platform on big-data tooling.
Shipped core platform modules and UI/testing frameworks in enterprise software.