A lot of this work is product work. Finding the real problem, choosing what not to build, and then running the program that gets it adopted — usually without a PM or a TPM in the room.
- Problem definition
- Reframed Meta's privacy review from a bespoke-per-review problem into a template problem — consolidating hundreds of thousands of mitigations into ~10,000 reusable primitives. The cycle time fell from 55 days to under 5 because the model changed, not because the code got faster.
- Driving adoption
- Built @trello/aidev solo, alongside regular delivery, then got the Platform team to adopt it and make it the foundation of Atlassian's AI SDLC workstream. Championed and led company-wide GraphQL adoption at Cvent, reshaping frontend API strategy.
- Measuring the thing
- Designed the analytics suite tracking 80+ operational metrics for reuse and accuracy at Meta, and automated evidence collection so 65% of requirements verify programmatically. He instruments what he ships and argues from the numbers.
- Cross-functional delivery
- Technical lead on a compliance system gating ~30,000 product areas, operating with Product, Design, Privacy, and Legal against tight deadlines. Led the full visual refresh of Trello's Planner end to end. The privacy-review overhaul had visibility up to Zuckerberg.
- End-to-end ownership
- atrium is the clean example: he set the roadmap, designed it, wrote the Rust and the React, wrote the docs, ran the releases, and supported the users — and those users now want to pay for it.
- Multiplying other people
- Mentored 13 engineers across two scrum teams at Cvent and founded re:Render, an internal UI education initiative. Ran biweekly knowledge-sharing sessions for 30+ engineers at Meta. The team grew from 10 to 80 engineers around this work.