AI-native product manager who ships the products I spec.
Most PMs write about AI. I build with it. I go from problem β prototype β deployed product on my own, with enough engineering depth to be dangerous and enough product sense to know what's worth building.
- π― Product: find a real, painful problem, scope the smallest thing that solves it, ship, iterate on usage.
- π€ Build with AI: RAG pipelines, LLM tool-calling, agentic workflows, and Model Context Protocol (MCP) apps. Not slideware, running code.
- π’ Ship end-to-end: React/TypeScript front end, Postgres + vector search back end, auth, payments, and cloud deploys.
- π€ Work with engineers as a peer: I read the codebase, open the PR, and speak the same language.
An AI assistant for British Columbia housing law. Tenants and landlords ask plain-language questions and get grounded, cited answers about their rights: rent increases, notices, deposits, renovictions.
π Live: rentadvise.ca Β β’Β also available as an app inside Claude and ChatGPT
What's under the hood:
- RAG over real law: embeddings + Postgres vector search over BC tenancy legislation and RTB forms, refreshed by nightly data pipelines (PDF scraping, OCR, Reddit ingestion).
- MCP server + custom widgets: the same assistant runs natively inside Claude and ChatGPT, with interactive UI components.
- Production-grade: Google/Facebook OAuth, Stripe billing, rate limiting, e2e tests, deployed on Railway with cron jobs.
I owned the product, the design calls, and most of the engineering, built solo with AI pair-programming (Claude Code, GPT). Being AI-native isn't a talking point for me; it's how this got shipped.
Stack: TypeScript Β· React Β· Node/Express Β· PostgreSQL + pgvector Β· Drizzle Β· OpenAI Β· MCP Β· Stripe Β· Railway
Beyond RentAdvise, here's what else I've shipped and put live:
| Project | Links | What it is |
|---|---|---|
| fly-with | π Live-demo π» Source |
Turn layovers into adventures β an iOS app for smarter long-haul stopovers. |
| pay-agent | π Live-demo π» Source |
POC agent that can use gift cards to make payments on a user's behalf. |
| Physical AI Deployment Command Center | π Live-demo π» Source |
Product OS for evaluating physical AI deployment readiness β safety gates, intervention metrics, failure taxonomies, customer ROI. |
| Collection Integrity CI | π Live-demo π» Source |
Local-first, offline QA/CI for museum collection data β deterministic rule engine, evidence-backed findings. |
- Ship small, learn from real usage, then decide what's next.
- Technical depth is a PM superpower: I can estimate what's cheap vs. expensive, and I can prototype the thing instead of arguing about it.
- The best spec is a working demo.
πΌ LinkedIn Β β’Β π rentadvise.ca