Healthcare. Compliance-heavy infrastructure. Multi-agent platforms that have to be reliable, auditable, and humane — not just impressive in a demo.
Currently: Principal Engineer at Watch Our Own, a HIPAA-compliant FHIR-native platform reducing 30-day hospital readmissions. Co-founder of Aegis Tech Ventures, where we apply the same patterns to other teams that need real AI outcomes, not slideware.
I started in accounting at KPMG and EisnerAmper — which sounds unrelated until you realize it taught me how regulated systems actually work, and what happens when the data underneath them is wrong. I built my own house on the Jersey Shore — framed it, wired it, finished it — because I wanted to know whether I could. I'm largely self-taught as an engineer, came up through a bootcamp, and have spent the last decade building backend and cloud infrastructure. The last few years have been AI: multi-agent systems, agent memory, ACL and HITL design, the FHIR layer underneath it all.
The thread is the same across all of it. I like building things that have to actually work — physical or digital, doesn't matter. The fun part is the bridge between them.
Why healthcare AI, specifically: my daughter was born at 29 weeks. The NICU experience was its own kind of clarity — information everywhere, and nowhere, at the same time. We'd come back in from a few hours away and find that the night shift hadn't reliably passed their observations and concerns to the morning shift. The people directly responsible for keeping our daughter alive weren't always communicating with the next set of people responsible for keeping her alive. That's the problem I've been working on ever since at Watch Our Own. Care coordination is a systems problem, not a clinical knowledge problem. The people around someone in care don't reliably share what they're seeing, and the systems built to help them don't bridge that gap either. AI is finally the layer that can fix that, if you build it carefully.
What I do day-to-day: I sit between clinical advisors who surface real problems, my co-founder Gian who owns the infrastructure, and the agent systems we're building on top. I write the application architecture from scratch, design the ACL and HITL surfaces, run the spec-to-implementation loop with Claude Code, and coordinate with the frontend teams across platforms to make sure the whole thing ships. I'm the person who makes the translation from clinical problem to working agent system reliable.
What I care about technically: design and data first. How information needs to live before you decide what the interface does. Spec-driven implementation over prompt-driven. Cloud-agnostic infrastructure so the patterns travel. HITL as a load-bearing layer, not a checkbox. Eval design owned by humans, not the model grading itself.
If you're trying to ship real AI into a regulated, high-stakes domain — and you're tired of demos that don't survive contact with production — that's the work I do.
HIPAA-compliant, FHIR-native multi-agent platform built to reduce 30-day hospital readmissions. The bet: readmissions aren't a clinical knowledge problem, they're a coordination problem. AI agents do the care-group coordination on a Medplum-backed clinical data vault. The consumer/caregiver layer is already built, which is what most clinical AI platforms are missing. Strategic focus now: hospitals and health systems.
Co-founded with Giancarlo Paolillo. We deliver production AI outcomes. The patterns we've proven at WOO — HIPAA posturing, FHIR-native data, multi-agent orchestration, cloud-agnostic deployment, HITL surfaces — applied to clients who need AI that ships, not AI that demos.
Best for: production AI infrastructure work, multi-agent system design, healthcare-adjacent AI, fractional principal engineering, or technical advisory.