Ethan Yan

AI Engineer · Agent Builder

Ethan Yan

I build AI agents that earn their place.

Engineer working on agentic systems — the planning, memory, and tool-use that make models genuinely useful. Ten years from backend to full-stack to AI. I care about software that feels calm, fast, and honest.

Building agent systems in public · 2026

Selected work

HR Agent A production HR / recruiting agent that writes code in a sandbox to call primitives (code-action), handling talent workflows end to end. AI Agent · Code-action · LangGraph
pdf2md A document-parsing pipeline that turns messy PDFs into clean, structured markdown for downstream LLM use. Python · PDF · LLM
Smart Building Report A three-stage AI construction-report pipeline (image/task recognition → unit reports → builder reports) over Java + Python + RabbitMQ + an LLM gateway. AI Pipeline · Java · Python · RabbitMQ
Model gateway & observability A LiteLLM-based, OpenAI-compatible gateway unifying GPT / Gemini / Claude with routing, retries, fallback, and caching — paired with Langfuse for end-to-end call tracing and cost observability. LiteLLM · Langfuse · LLM Infra · Observability
LLM evaluation platform Multi-model × multi-prompt evaluation with LLM-as-Judge scoring, batch runs, and resume-on-failure — built on the same LiteLLM + Langfuse stack. Eval · LLM-as-Judge · LiteLLM · Langfuse
pingcode-mcp An MCP server that bridges the team PingCode project-management tooling to agents. MCP · TypeScript
Internet hospital Backend for online consultation and appointment booking — interface development, business logic, and on-call stability work. Java · Healthcare · High-availability
Distributed bank custody system A high-concurrency financial transaction platform across upstream banks and downstream businesses — 30M+ transactions/day at 99.99% uptime, with JVM/DB/cache tuning lifting throughput ~80%. Java · Spring Boot · Redis · RabbitMQ · Oracle

Recent writing

All posts →
Inside Kimi Code: a teardown of its agent engine Don't put symptoms in the prompt The verifier is the hard part

About

I'm Ethan — an AI agent engineer who got curious about what happens when you give a language model hands. These days I spend most of my time on the unglamorous middle layer of agents: the retries, the guardrails, the memory that keeps a system coherent past the first prompt. More about me →