I build AI agents, RAG assistants, and custom MCP servers that connect to your real tools and data — plus edge-AI on ESP32. Full-stack engineer & CTO, working remotely with clients worldwide.
From an idea to something that ships and keeps working.
Agents that use your tools and data (RAG, tool/function calling) — not just chat.
Connect Claude, Cursor, or your own AI to your APIs, databases and internal tools, securely.
Turn PDFs, scans and forms into clean structured data (OCR + LLM extraction + workflows).
Pipelines, dashboards, extraction, matching and recommendation over your business data.
Smart firmware on ESP32 (ESP-IDF / Arduino / TinyGo), with on-device AI.
Production systems in Python/FastAPI, Go, and Next.js.
Real systems — with screenshots, code and live deployments. Live demos are password-protected; access available on request.
Turns a flood of public tenders into ranked, actionable opportunities.
An AI platform that crawls public procurement data, uses an LLM to extract and structure each notice, then matches and recommends opportunities to a company by capability — with match scores, risk flags, and a natural-language search. Full pipeline: crawlers → LLM extraction → pgvector matching → recommendation.

Client project — designed and built a capital-markets research & reporting web platform to a financial-services client's requirements, plus a customized second deployment. Mock data only.



End-to-end document verification using computer vision (YOLO + Siamese-network embedding) with a backend pipeline and human-review workflow.

Production-grade MCP servers (Go) connecting AI assistants (Claude, Cursor) to real tools via OAuth2 — minimal, secure, cross-platform.

Turns messy requests from enterprise chat into structured data with LLM extraction + human review — then renders print-ready documents from templates. Phase-1 prototype.

Restyled the React frontend, adapted the API, and containerized & deployed an open-source stock-analysis platform — running in production with daily AI market reviews.
Built on the open-source ZhuLinsen/daily_stock_analysis — my work: frontend redesign, API adaptation, deployment.
I'm Yuexiang Yuan (Tom) — a full-stack engineer with 15+ years' experience and currently a CTO. I help businesses connect AI to their real systems: AI agents, RAG, and custom MCP servers that run in production. My background spans backend (Python / Go), DevOps, Web3, and embedded / edge-AI. I ship production-grade work with documentation, communicate clearly in writing, and offer ongoing support. Based in Nanjing, China — working remotely with clients worldwide.