At a dermatology clinic seeing 70 patients a day, I identified the documentation bottleneck, built an AI scribe against it, and got it featured as a customer case study on Omi. Same-day chart completion moved from 60% to 95%. Since then: six AI products shipped — four Omi agents (2,200+ users) and two native iOS apps in public beta on TestFlight. I own every step: customer interview → prompt design → native module → deploy → iterate.
Designed and deployed an always-on AI scribe for a dermatology clinic seeing 70 patients a day. Structured prompts, EMR-compatible note generation, and a real workflow that clinicians actually use. Deployed 2024–2025; case study published November 2025.
Replaced manual charting with an AI workflow that records, transcribes, and generates structured HPI notes formatted to match the clinic's EZDERM EMR templates. HIPAA-compliant. Now used daily by the clinic's doctors, nurses, and medical assistants.
Full case study on omi.me ↗Our notes are way more accurate and complete — no longer dependent on someone's good or bad day. The AI gives us the same high-quality structure every time.
Two native iOS apps, architected and shipped to TestFlight. No developer-operated backend, no analytics pipeline, no tracking — privacy-first from the ground up.
Privacy-first AI assistant for iPhone and iPad. Apple Intelligence on-device by default, optional Ollama local or cloud. Conversation search, memories, document import, voice input, image attachments, and built-in tools.
Join the beta on TestFlight ↗
Private-by-default iOS AI assistant. Dual on-device inference (Apple Foundation Models + Ollama), 8 custom Swift modules, Live Activities, Vision OCR. Comprehensive test suite, ~5,000 LOC of native Swift.
Join the beta on TestFlight ↗
Most “React Native” apps lean on JavaScript wrappers. N.O.V.A. ships 8 custom Swift modules — real platform integration, not Expo plugins — plus a comprehensive test suite guarding every change. Source is private while I finish cleaning it up; happy to share during interviews.
novaai, backwards-compatible).web.ts shadows for every native module so the app builds on web for testingfunction getAIProviderStatus(
aiProvider: { provider?: AIProvider } | undefined,
ollamaConnectStatus: ConnectStatus,
fmIsAvailable: boolean,
) {
const selectedProvider = aiProvider?.provider ?? AIProvider.FOUNDATION_MODELS;
const isFoundationModelsSelected = selectedProvider === AIProvider.FOUNDATION_MODELS;
const isFoundationModelsUsable = isFoundationModelsSelected && fmIsAvailable;
const effectiveProvider = isFoundationModelsUsable
? AIProvider.FOUNDATION_MODELS
: AIProvider.OLLAMA;
const isFallbackActive =
isFoundationModelsSelected &&
!fmIsAvailable &&
ollamaConnectStatus === ConnectStatus.SUCCESSFUL;
return { selectedProvider, effectiveProvider, isFallbackActive };
}
Two clinical workflow tools for dermatology practices and two consumer AI assistants. Each one is live, installed, and getting used today.
Open to Founding Engineer, Forward Deployed Engineer, and senior Applied AI roles. Burlington, NC — remote or Triangle-area hybrid. The best projects are ones where a real user's workflow is on the line.