Full-stack engineer who ships · AI, backend, DevOps
Shipped products, client work, and systems in daily use, from AI on-device inference to booking platforms.
An AI facial skin-analysis and skincare-recommendation platform. I built it end-to-end: acne analysis, recommendations, auth and user-data flows. I engineered a custom JWT flow for LINE LIFF tokens and optimized a YOLOv11 model to TensorFlow Lite for on-device inference. When SaMD medical-device requirements surfaced, I re-platformed the MVP to LINE LIFF with zero rework of core logic, keeping launch on schedule.
Booking platform for a Karen community's eco-tourism, replacing a manual phone-and-Word process. One 4-step wizard serves activities and homestays, with webhook-confirmed payments and inventory held before pay to prevent double-booking.
Elderly-caregiver matching platform. A scalable modular-monolith backend with GraphQL, role-based access control, and secure API contracts protecting sensitive healthcare data, with a roadmap toward microservice extraction.
A full-stack request system used daily by librarians, staff, and students across CAMT, CMU. Migrated legacy auth to CMU OAuth SSO, built GitHub Actions CI/CD that cut deploy time 40%, and containerized services for 99%+ uptime.
Built a booking platform (Next.js, NestJS, Prisma/PostgreSQL, Beam) that replaced a manual reservation process, unifying activities and homestays into a single 4-step wizard with webhook-confirmed payments. Led community requirements gathering and implemented the frontend from the designer's Figma system.
Built Python automation pipelines with Papermill to shuffle datasets and run ML baselines, cutting manual data-prep time. Benchmarked local open-weight LLMs (Qwen2.5, Granite) on VRAM footprint, inference time, and tool-calling reliability to optimize the orchestration pipeline.
Architected a scalable modular-monolith backend with NestJS + GraphQL and a clear roadmap for microservice extraction. Designed and implemented role-based access control and secure API contracts protecting sensitive healthcare user data.
Built the product end-to-end in Flutter and Supabase, engineered a custom JWT verification flow for LINE LIFF tokens, and optimized a YOLOv11 model to TensorFlow Lite for on-device inference. Re-platformed the MVP to LINE LIFF for SaMD compliance and onboarded 164 users across three field deployments.
Developed and deployed a Book Request System used daily across CAMT (Node.js, MySQL, EJS), migrated legacy authentication to CMU OAuth SSO, built CI/CD with GitHub Actions that reduced deployment time 40%, and containerized services with Docker + Portainer to raise uptime above 99%.
I turn rough ideas into validated, production-ready products. I like wearing many hats: architecting backend infrastructure one day, training computer-vision models the next, shaping startup strategy after that.
Right now I'm founder and lead engineer at SkinNavi, working to replace skincare trial-and-error with precision computer vision and research-backed data. I care most about shipping things people actually use, and iterating fast once they do.
