I find the human signal inside complex systems — and turn it into products that work.

Human signal

Customers couldn't get useful answers out of the existing chatbot. Too often it simply replied that it couldn't help, and the conversation ended there.

Product move

Sole PM for a team of 10+ engineers and designers, concept through working prototype. Reworked retrieval and sub-agent routing against what 20+ user interviews said people were actually asking for.

Human signal

Agents hold dozens of customer conversations at once, each carrying its own history and paperwork, with no good way to keep that context straight.

Product move

A second 0-to-1 product, again as sole PM with 10+ engineers and designers: a co-pilot that assembles the context for each customer and drafts the next response for the agent to approve.

Human signal

Rebuilding an enterprise platform meant product owners spending weeks reading legacy documentation before they could write a single epic or user story.

Product move

Agentic tooling that reads the legacy documentation and drafts epics and user stories for a product owner to review, correct and keep — turning weeks of reading into an afternoon of editing.

Human signal

My HBS section had nowhere to see what was happening that week, no way to match faces to names, and no quick route to a sectionmate's career history.

Product move

A web app I designed and shipped myself, scoped from five interviews with sectionmates: one calendar, one directory, one place to find the person who has done the thing you're about to try.