I'm a systems thinker, a product builder, and a leader of engineering teams. I work in both directions: translating business goals down into technical strategy, and carrying ground truth back up so the goals get sharper. Understand the system end to end, pick the leverage points, build toward them, then measure whether it actually worked. At Meta I grew and led teams from 5 to ~30 engineers, and I treat the team itself as a system: clear goals, the right structure, and senior engineers who get visibly better under you. Fifteen years across Meta, consumer fintech, and two companies of my own — from data platforms at a trillion events a day to iPad point-of-sale in restaurants. Underneath all of it I'm a deep technologist and a tinkerer: when a tool disappoints me, I take it apart and build a better one. The projects further down this page are that half of the work.
I used data to cut harm on social platforms. At Meta I built the enforcement data systems behind content moderation on Facebook, Instagram, and WhatsApp: billions of signals a day across hate speech, misinformation, terrorism, and child safety; 15,000+ reviewers routed by queues my team designed; 95%+ proactive detection on key harm categories; a measured 10% drop in overall harm prevalence. In plain terms: less harmful content reached users, and the numbers behind that claim were instrumented well enough to defend in front of the whole company.
Security was the same problem with a harder constraint. Leading ~30 engineers in Meta's security org, I turned vague posture into measurable goals and hit them: encryption at rest from 62% to 90%+, MFA enforcement to 100%, vulnerability remediation inside SLA at 85%, residual access risk down 9% quarter over quarter — while access friction fell 5% month over month. Users and privacy stayed protected and the business kept shipping; getting both moving in the right direction at the same time was the actual job.
I also evolved the architecture of one of the largest datasets in the world. Project Merlin unified viewability logging across Meta's surfaces — a trillion events a day, one protocol replacing dozens of inconsistent client implementations, 15–70% redundant event load eliminated — and became the measurement foundation for ranking, delivery, and $185M+ a day in ads attribution. Migrating live systems at that scale without breaking them is where most of my judgment about trade-offs comes from.
