Resume
Yeqiu (Jason) He — AI Systems Engineer: agentic systems, RAG, LLM integration.
Summary
I spent five years building process automation for enterprise IT — unglamorous work that taught me how production systems actually live: they ship, they break, someone has to notice, and someone has to fix them. When LLMs arrived I was too curious to watch from the side, so I learned the way I always have — by building real things and running them myself: a small fleet of agents on a Raspberry Pi that helps organize my day, a retrieval system a working lawyer relies on, this site. Most of what I know about agentic systems and RAG comes from operating these for months and studying how they fail — not from any course.
Selected work
- An eight-agent fleet — eight agents I run at home, built largely to find out what breaks when agents must remember and cooperate over months. They handle my daily briefings, my publishing flow, and much of this site's upkeep. The memory architecture and inter-agent protocol are mine; the lessons were mostly taught by the failures.
- Meinrag — a RAG system a practicing lawyer uses daily over a 3,300-document corpus. The part I care most about is not the retrieval stack but a principle: every answer must cite its source, down to the exact box on the PDF page — a system someone relies on professionally should show its work.
- Five smaller builds — music discovery, board-game engines, language tools — each one an excuse to understand something; documented on the blog and YouTube.
Skills
- AI / ML
- LLM integration & orchestration (Claude, OpenAI, Codex) · RAG (FAISS, embeddings, re-ranking, hybrid BM25/dense) · MCP servers · multi-agent systems · model calibration & evaluation · vector search · Monte Carlo simulation
- Programming
- Python · TypeScript / JavaScript · C# · VB.NET · SQL · Fortran
- Backend / Data
- FastAPI · SQLAlchemy · PostgreSQL · MySQL · SQLite · ETL / data-pipeline design · REST APIs
- Automation
- UiPath · PowerShell · Windows automation · process optimisation
- Infrastructure
- Docker · Linux · Git · self-hosted ops (Raspberry Pi / NAS / VPS)
- Frontend
- Vue.js · React · Nuxt · Tailwind · Astro
Languages
English (full professional proficiency) · Mandarin Chinese (native) · German (basic)
In place of a cover letter: I treat AI as a tool for curiosity, not a substitute for thinking. Most of what is on this site exists because I wanted to understand something, and building is how I read. I am aware of how much I don't know yet — the projects above are my way of finding out.
The full résumé — contact details, work rights and availability — goes out privately per application. Reach me via GitHub.