AI product manager. I build agentic systems, generative AI, and production ML for the enterprise — where the hard problems are rarely the models, and almost always trust, evaluation, and what an organization will actually adopt.

Everything here is working code for a decision I've actually had to make:

• ai-product-decisions — PM judgment, made runnable: evals, workflow-vs-agent arithmetic, failure analysis
• agentic-copilot — a stateful, tool-using conversational agent (LangGraph + MCP), end to end
• document-ai-bench — an open benchmark for LLMs on long, structured, high-stakes documents
• How to make AI products — my free 24-unit applied GenAI curriculum, no sign-up

Longer thoughts live at carlosarivero.com. Say hello on LinkedIn.

Day job: leading product for teams of humans and AI. One side hallucinates facts, the other hallucinates deadlines. I manage both.