I build AI agents and the tools around them. I'm curious about what happens between a model's answer and a finished task: context, tools, human decisions, and the evidence that it worked.
Currently deep in coding agents, with one foot in runtimes and the other in product delivery. Still happiest when I can look underneath the abstraction.
Open to AI Agent Engineering roles, including remote. HKU Computer Science MSc requirements completed; degree conferral expected November 2026.
A coding agent you can look inside. My main workshop for tool approvals, recoverable sessions, context compaction, Skills, MCP, and sub-agent coordination. Built in TypeScript, all the way to the terminal experience.
I also run paired repository-repair evaluations and check graders against raw test logs. Experiments, corrections, and results.
A Rust CLI that makes Skills easier to manage across 40+ agent tools, locally and in Docker. Declarative installs, bounded concurrency, and a doctor / repair loop when things drift. Built through a human-led, multi-agent workflow.
A team-built quantitative research workbench. My contributions connect natural-language screening, refresh-safe streaming progress, and deterministic backtesting across Next.js, FastAPI, and storage.
TypeScript for agents and interfaces. Python for services and evaluations. Rust for tooling. Java / Spring Boot for backends; Kotlin for coroutine pipelines.
A little further under the hood
- Agent systems: tool calling, context and session management, sub-agents, MCP, Skills, human approvals.
- Product delivery: React / Next.js, FastAPI, Java / Spring Boot, SSE, SQLite, PostgreSQL, Redis.
- Systems: Linux, Docker, CI/CD, async runtimes, concurrency control.
- Before this: at Guangdong Genius Technology, I worked on Java / Spring Boot backend services and built a Kotlin coroutine review pipeline handling roughly one million new questions every two weeks.
Repository counts and star totals cover personally owned public non-fork repositories. Project cards show each project's own stars. Data snapshot.
If you're building coding agents, developer tools,
or a team around them, I'd love to compare notes.
