AI Native Product Manager
I move the work people do by hand into AI agents — and prove it with products that ship.
Five years in Web3 as a BD / PM. I led global go-to-market end to end, then moved beyond specs and started building the products my team actually runs on — AI agents, data dashboards, content pipelines. I'm not an engineer; I build to validate product decisions, and the build is the proof.
Everything below is public — and most of it is running live right now.
🟢 canton-hub · live ↗
Real-time dashboard for Canton Network. Collects scattered external sources on a schedule, serves them to the frontend over REST + SSE. FastAPI Next.js
🟢 polymarket-community-calendar · live ↗
Polymarket prediction-market data, visualized as a calendar and timeline. React TypeScript
signal-to-story
A market-signal → short-form-content pipeline: script generation, capital-markets compliance review, and a human approval gate, wired into one flow. Runs keyless via the local Claude CLI. Python Claude
claude-code-harness
A non-engineer's Claude Code harness, public edition — the docs → plugin → automation onboarding framework, a 3-model judge-panel scorer, and ops notes from running daily LLM bots. Claude Code Python
canton-telegram-bot · kospi-morning-bot
Daily report bots running unattended on launchd. The KOSPI bot grades its own morning predictions against actual closing prices, every trading day. Python LLM ops
krx-ai-bot
Event-driven autonomous trading daemon for KRX. Claude reads the market regime; rule-based risk limits do the gatekeeping. Paper-trading by default. Python asyncio
- Moved repetitive team work into 18 in-house AI agents & skills, then onboarded the team so it stuck as shared infrastructure. The public half lives in claude-code-harness.
- Turn scattered data into live dashboards, not slide decks.
- Ship LLM / agentic workflows myself with Claude Code — tools, not demos.
No streak counters or commit graphs here — the live products above are the metric.