AI Engineer: Code-First, Depth-Driven Expertise | Generated by AI
You’re Zhiwei (lzwjava) — here’s what I know about you from your custom instructions:
Identity
- Software engineer, 12 years of experience across iOS, Android, frontend, backend, and AI
- Self-taught, university dropout, learn by building
- Built Fun Live (30k users, 3M CNY revenue)
- Currently AI Engineer at a global bank (Singapore/Hong Kong financial systems background), ranked top 6% globally in AI assistant usage
Technical Profile
- Train models from scratch (GPT-2 on H200/H100 GPUs)
- Consume ~1.5B LLM tokens/year
- Build CLI agents:
ww,iclaw,zz - Public knowledge base: ~8,000 AI answer notes at lzwjava.github.io/notes-en
- ~400 technical blog posts at lzwjava.github.io
- Stack: macOS, terminal-first, Python primary, OpenRouter, Jekyll, GitHub Actions
Idols
Yin Wang, Andrej Karpathy, Wenfeng Liang, Greg Brockman
Long-term Goal
Full transition into AI engineering — model training, agent architectures, LLM internals, AI-native tooling
How I Should Answer You
- Karpathy-style depth — first principles, math when it matters, one layer deeper than expected
- Code-first — minimal implementations over prose
- AI-native bias — LLM/agent approaches over traditional patterns
- English only — even if you write in Chinese
- Search your notes first — don’t re-explain what you’ve already covered
- No shallow “it depends”, no corporate-speak, no beginner hand-holding