Tai
Huynh
I build AI-powered features that ship — RAG pipelines, document intelligence, semantic search, MCP servers.
Every project security-reviewed, tested, and deployable.
chapters i — iii
A working index
of recent work.
Web Research MCP
An MCP server giving any AI assistant safe web access.
Built for the real world.
Production-grade isn't a marketing phrase. It means features that can be deployed by someone else's team, survive a security review, and keep running at 3 AM without a phone call.
All 11 projects are security-reviewed and built on a consistent stack so patterns are reusable and integration is predictable.
“Eleven projects · two hundred eleven tests · one stack, one craft.”
Ship to production
Every project runs in a real environment with real constraints. No toys, no demos. Auth, rate-limiting, error handling, observability — in by default.
Security by default
SSRF defense, prompt injection mitigation, input validation via zod, secrets-free build. Each repo ships with a SECURITY.md — not as an afterthought.
Tested and observable
211 passing Vitest tests across the portfolio. Logging and structured errors everywhere. CI runs on every push.
One stack, deep craft
Consistent tooling across all 11 projects means transferable context, faster debugging, and a codebase that reads the same way every time.
Let's build
something.
Available for freelance AI work — RAG, structured extraction, embeddable widgets, MCP servers, internal tools. Typical turnaround 2–7 days.