I'm a PM on the new products team at GoFundMe, where I've shipped AI-powered features end-to-end: a multi-step generation pipeline for user-facing content, prompt systems tested against real user inputs, and an eval system that holds quality steady in production. I work in the actual mechanics of building with AI: scoping what an agent should do, structuring the pipeline, testing it against real inputs. And I know how this works inside a real org, where shipping AI means coordinating handoffs across Product, Design, and Eng. That's where most enterprise AI efforts stall. What sets me apart is that I build outside of work too. I've shipped several AI products solo with agentic coding tools: a fitness app, an AI image generator, a personal CRM, etc. I run a tight loop with AI at every step: write the spec, break it into tasks, execute with coding agents, then capture what worked and what broke in a file the agent reads next time. Each loop ships faster than the last. When I coach, I'm teaching the loop I run myself.