Sr. AI PM building agentic systems, semantic data models, and the strategy to ship them.
| Project | What it does | Stack |
|---|---|---|
| semantic-model-kit | Governed business context for a warehouse so LLM agents answer correctly. Compiles to Snowflake semantic views, Open Semantic Interchange, and dbt MetricFlow; serves metrics over MCP; measures the accuracy gain with a three-mode eval. | Python, DuckDB, MCP |
| closer-ai | An AI phone agent for small businesses: answers, books, follows up, attributes every lead, and reports. Runs end to end with no API keys. | Python, FastAPI, React, OpenAI Realtime, Twilio |
| agent-evals | A PM-authored scorecard and CLI that prove an agent works, not just that it demos well. | Python, LLM-as-judge |
| enterprise-ai-playbook | The strategy and governance artifacts a company needs to roll out AI assistants people trust and use. | Markdown |
| papaya | A Spanish-learning game for kids ages 5 to 10. Lessons earn play time; a Claude-voiced mascot is optional. | TypeScript, React, Vercel |
| product-feedback-mcp | An MCP server that lets an agent search, cluster, and triage product feedback against real data. | Python, MCP |
| rag-starter | Retrieval from a simple chain to hybrid search, with an eval set that shows when each wins. | Python, ChromaDB |
Also: claude-skills, self-healing-browser-agent, voice-agent-starter, llm-finetuning-recipes.
Savage Companies, Sr. Product Manager, Applied AI. Own the AI product strategy and roadmap for a multi-company operating group. Shipped an agentic risk-reporting workflow that cut turnaround roughly 80% across 7 teams, and a self-serve data platform with text-to-SQL now used by 25+ teams at roughly 70% time savings. Current focus: semantic modeling so agents answer from certified definitions.
Microsoft, while working as a Product Manager on Azure AI Foundry. Owned memory partitioning reliability, prioritized scaling decisions from platform telemetry, and led open model validation for the Llama and DeepSeek families ahead of general availability.
Closer AI, co-founder. Built the retrieval pipeline and fine-tuned Llama-3-8B for live sales calls, then designed the move to direct SIP into a realtime speech model. TriagePoint.AI, co-founder, a Bayesian ML triage product that won Best in Medicine. Directed the Lassonde DevLab incubator, 20+ AI-first startup teams.
Earlier: Senior PM at GardaWorld across a consumer app and an enterprise B2B product; product roles at Ground and Up (co-founder), Royal Anne, and Liberty Smart Home. 7+ years of product management.
MBA and MS in Computer Information Science, University of Utah.
Happy to talk about agents, evals, data products, or enterprise AI adoption. LinkedIn or email above.