$ whoami --verbosename: Chun Ki Lam (a.k.a. Jacky)
role: Technical leader — infrastructure, cloud, backend, AI
xp: 15+ years 🎮 (infra → cloud → backend → AI)
location: Hong Kong 🇭🇰 (UTC+8)
langs: Cantonese 🗣️ | Mandarin | English
studying: M.Sc. Big Data Analytics, Univ. of Liverpool 🎓 (2026)
currently: teaching LLMs to remember things, and servers to behaveI've spent a decade making infrastructure boring — in the best way. Racks, hypervisors, multi-region clouds, then APIs, then AI systems that actually ship. If it needs to run at 3am without waking anyone up, that's my happy place. 😴
| Project | What it is | Stack | |
|---|---|---|---|
| 🎮 | FPCHK | AI-enabled esports platform with 4,000+ users. OCR reads match screenshots so humans don't have to. | Django PostgreSQL OCR Telegram |
| 🌳 | Treehole | Safety-aware conversational AI with persistent memory + multi-level risk escalation and provider fallback. | FastAPI pgvector LLM |
| 💡 | Lumi | Voice-first multilingual AI coach (粵語 / 普通話 / English) using Socratic prompting + TTS. | Flutter LLM TTS |
| 📚 | Dictation Pal | Offline-first learning app for kids, with parent controls. Works on a plane. |
Flutter SQLite/Drift OCR |
- 🔭 Building AI systems that remember, reason, and fail gracefully
- 🌱 Finishing my M.Sc. in Big Data Analytics
- 🧪 Poking at vector retrieval, prompt orchestration, and agent workflows
- 💬 Ask me about: OpenStack war stories, scaling Django, or why your VM is slow
- 🤖 Building agentic development workflows with Codex, Claude Code, OpenCode, and MCP
- 🧪 Exploring tool use, multi-agent orchestration, evaluation, and reliable LLM systems
- ⚡ Fun fact: my first "production system" tracked aircraft. Uptime was non-negotiable.
