About
Most AI product failures aren't model failures — they're context failures. The model gets thin inputs, produces thin outputs, and the team blames the technology. I focus on the part before the prompt: what context does the user need to give the system, how do you make that easy, and how do you know the output is actually useful.
That's been the core of my work at HubSpot: building the context infrastructure that makes AI-generated campaign assets accurate and on-brand, not just fast. It's what I brought to Orum before that, where getting the context architecture right turned meeting summaries from "kind of useful" into something AEs relied on every day.
I've built products across retirement fintech, wealth management, sales intelligence, and marketing technology. The common thread: reducing friction between user intent and system output, at scale.
