Recruiter turned builder. I've spent years inside the recruiting problem — I know exactly where it breaks down. Now I'm building the tools I always wished existed.
icebreaker.build is live and running. No co-founder. No dev team. Just deep domain knowledge, product instinct, and Claude Code as my engineering partner.
| Product | What it does | Status |
|---|---|---|
| icebreaker | AI recruiting email writer — upload a resume, get a personalized outreach email in 30s | 🟢 Live |
| icebreaker-Pro | Batch recruiting outreach — CSV import, AI generate, review, SMTP send | 🟢 Live |
| Talent-Flow | Lightweight ATS with Kanban pipeline, stage history, and candidate tracking | 🔨 In progress |
| researcher-mapper | AI Researcher Discovery Platform | 🟢 Live |
| github-digger | Find Chinese contributors from GitHub for talent recruiting | 🔒 Private beta |
I don't write code. I bring the thing most engineers don't have: deep knowledge of the problem from the inside.
Every line of code in these repos was written by AI. I make the product decisions, define the UX, and run relentless quality iterations until it's right.
Domain expertise + AI as engineering partner. That's the model I'm testing — and it's working.
AI / LLM engineers who want to work on real problems, not toy demos.
The recruiting outreach problem is harder than it looks:
- Making 500 emails feel like 500 individual letters, not 500 variations of one template
- Cross-lingual generation that preserves tone and nuance
- Candidate profile understanding that goes beyond keyword matching
If that sounds technically interesting to you — let's talk.
📍 Shanghai