Founder, Researcher, Engineer
Drawn to what doesn’t exist yet; and wired to make it real.
Founding engineer who ships 0→1 across full-stack products and integration-heavy healthcare backends. Launched Xuman.AI on the App Store in ~3 months with a team of two, owning mobile, backend, infrastructure, payments, and real-time video. Built production EHR integrations (FHIR R4, Canvas Medical) and clinical data pipelines from scratch. Strong at rapid iteration, production reliability, and translating ambiguous requirements into working systems.
Product-driven software engineering generalist. Currently building production EHR integrations for a stealth healthcare startup while continuing to lead Xuman.AI. Founded Style.AI to bring AI-powered fashion intelligence to the real world. Published researcher at AAAI and IEEE on LLM-based IoT security and ML for fusion energy. Hackathon winner (SF Hacks 2024 — Best GenAI Hack). Refounded and led the AI Club at SF State as President.
Stealth Startup — Freelance Healthcare Integration Engineer
Jan 2026 – Present · Remote
Architected and built production EHR integration with Canvas Medical (FHIR R4): OAuth2 authentication, patient CRUD, appointment scheduling, insurance coverage creation, and real-time eligibility verification via Claim.MD clearinghouse.
Developed clinical event plugins (Python) with HMAC-signed webhook handling, and payor normalization layer mapping consumer insurance names to FHIR Organization references for eligibility workflows.
Led PHI architecture refactoring, removed patient demographics from application database, fetching on-demand from EHR to eliminate HIPAA-compliant hosting overhead.
Marketplace with Agentic AI Workflows
Took Xuman from concept to production in ~3 months with a team of two. React Native (Expo) mobile client, NestJS microservices, Postgres/Prisma, Redis caching, LiveKit/WebRTC real-time video, Stripe Connect payments, and Azure deployments. iOS live on the App Store.
AI-Powered Fashion Intelligence
Full-stack AI wardrobe assistant (React Native, FastAPI, PyTorch) that scans clothing via computer vision, generates personalized outfit recommendations, and uses an active learning pipeline trained on 52K+ images. 200+ users in the first month.
Researcher
Feb 2024 – June 2025 · San Francisco, CA
Developed ML surrogate models for predicting plasma behavior in fusion tokamak devices; collaborated across research stakeholders.
LLM/RAG-based IoT attack detection using feature ranking and knowledge-base prompting; evaluated on public IoT datasets.
ML Surrogates for Fusion Tokamak Plasma Prediction
ML surrogate models for fusion tokamak plasma prediction. Multi-institutional effort with MIT, Princeton Plasma Physics Lab, and LBNL. Increased efficiency by 25%. Collaborated across research stakeholders including MIT, Princeton Plasma Physics Lab, and LBNL. Ran large-scale training on NERSC Perlmutter HPC clusters.
LLM / RAG-Based IoT Security on Edge Devices
LLM/RAG-based IoT attack detection with feature ranking. Accepted at AAAI Spring Symposium 2025 and IEEE DSAA-SF 2024. Evaluated on public IoT datasets using feature ranking and knowledge-base prompting to enable efficient on-device attack classification without cloud dependency.
Tinkering with ideas outside the flagship work.
Published at AAAI, IEEE, APS. SF Hacks winner.
Let's talk
Open to full-time roles, interesting collaborations, and conversations about building great products.
PDF · Updated 2026