I build AI that keeps your most sensitive data out of someone else's cloud.

Fifteen years across healthcare, fintech, security, and defense taught me the same lesson everywhere: the harder the data, the more it matters where it lives and who can touch it. In healthcare that is non-negotiable. Patient data cannot casually leave the owner's control, and most AI still assumes it can.

So I build the other way. Sensitive data stays local — on-device, on-prem, or in the client's own cloud — and only what is safe to share ever leaves. Frontier models in the cloud still have their place; they just work on de-identified data, never the raw record. Privacy by architecture, not by policy.

Right now that means an ambient clinical scribe that captures and de-identifies on the device before anything moves, built for clinics in Mexico and other markets that cloud-first, English-first tools ignore. The same instincts carry across everything I have built — securing sensitive systems, financial infrastructure, and cleared defense work — but healthcare is where it matters most, and where I spend my time.

01 Sovereign AI

Local, on-prem, and open-weight models that run where your data already lives — on-device, on-prem, or in your own cloud — so patient records, code, and IP never leave the owner's control. Frontier models still have a place; they just work on de-identified data. Privacy by architecture, not by policy.

02 Clinical & healthcare AI

HIPAA, EHR, claims, and clinical workflows, plus multilingual local clinical AI for markets that English-first tools ignore.

03 AI security & governance

Making AI both powerful and containable: controlling what models can access and keeping sensitive data from crossing the wrong boundary. Patent holder in AI governance.