"From Healthcare to HealthTech: Building production-grade AI systems with ICU-level discipline."
I don't just use AI; I orchestrate it. In the last 90 days, I transitioned from a clinical professional to a technical builder by shipping over 30 projects, culminating in a full-scale Clinical Decision Support Platform. I specialize in high-velocity implementation, turning complex clinical requirements into "zero-fail" code.
A full-stack clinical intelligence platform built to simulate hospital-grade decision support.
Live Demo | Architecture Deep Dive
- Resilient AI Orchestration: Implemented a Model Fallback Chain (Gemini 2.5 → Flash Lite → Gemma 3) to ensure 99.9% uptime despite API quotas or latency issues.
- Clinical RAG Pipeline: Developed a semantic search engine using Supabase pgvector and 768-dim embeddings with custom chunking logic for unstructured medical notes.
- Interoperability: Built a FHIR R4 Explorer that pulls live data from HAPI FHIR servers directly into the AI workbench.
- Security & Ops: Integrated Zod-validated request guarding, Upstash Redis rate-limiting, and a structured Audit Logging system for HIPAA-aware traceability.
- Medical Grounding: Live PubMed API integration to cross-reference AI analysis with peer-reviewed literature.
- Core: Next.js 15/16 (App Router), TypeScript, React 19.
- AI/Data: Vercel AI SDK, Google Gemini/Gemma, Supabase (pgvector).
- Infrastructure: Redis (Upstash), GitHub Actions (CI/CD), Vitest (Integration Testing).
- Domain: HL7 FHIR R4, PubMed eUtils, Clinical Workflow Automation.
As a former Special Forces Leader and ICU Nurse, I know that in clinical environments, "it mostly works" isn't good enough. I build software that prioritizes:
- Observability: Real-time metrics and latency benchmarking.
- Explainability: "Show-your-work" panels with retrieved evidence chunks.
- Reliability: Fail-fast environment validation and robust error handling.
LinkedIn | Portfolio | Status: US Citizen | NC-Based | Available for HealthTech Engineering Roles