I build AI systems that make consequential decisions about people, and I measure whether they get them right. At Meta I launched an LLM-powered moderation product running at 94% human-validated precision, red-teamed Meta AI before its Threads launch, and grew Community Notes on Threads 18× while cutting its time-to-consensus 4×. Five years and 20+ launches in trust & safety, most of it spent deciding where automated judgment should and should not be trusted.
Experience
Meta
New York, NY · 2021 – Present- Launched an LLM-powered moderation and enforcement product from zero: set the precision bar, validated it against human-labeled data at 94% precision, and shipped automated enforcement that has actioned 310,000+ accounts while cutting time from report to action from days to under an hour. Productized it into shared infrastructure now used by child-safety and harmful-link teams.
- Red-teamed Meta AI before its Threads launch, probing for adversarial prompts, jailbreaks, and misuse vectors, then defined and shipped the safety controls that followed.
- Led Community Notes on Threads from zero to launch across iOS, Android, and Web, treating rater quality as the product: grew daily active raters 18×, lifted 14-day rater retention over 70%, cut time-to-consensus 4×, and expanded note coverage 20×.
- Owned the response to a platform-wide impersonation crisis across 10 launches in one half: impersonation reports down over 40%, detection nearly 4× faster, and protected-identity coverage expanded by three orders of magnitude.
- Own roadmap and prioritization across 4 platform teams plus policy, legal, and operations partners; mentor engineers and lead weekly privacy and security review.
- Engineering lead for election integrity transparency: shipped state-linked account labeling on Instagram across backend, iOS, Android, and Web for the Philippines, Brazil, and US midterm elections, covering accounts responsible for billions of civic content views.
- Engineering lead for Threads-only accounts, shipping standalone account creation, login, recovery, and settings on iOS and Android under a fixed regulatory deadline.
- Built Instagram's identity-protection foundations, expanding About This Account eligibility 7× to tens of millions of users (+70% daily views) and automating enrollment of impersonation targets into protection.
Earlier Experience
Coinbase · SWE Intern, Coinbase Prime: Institutional trading features in Go/React making investment-vehicle setup 10× faster; FIDO2 authentication.
2021Amazon · SWE Intern, AWS: Aurora Serverless auto-scaling and a DNS library spanning 100K+ database clusters.
2020Apple · SWE Intern, Core Siri: Microservices parallelizing Siri workflows, improving testing times 300%. Shopify · Payments: split-payments to 70K+ merchants on POS.
2018 – 2019Skills
- AI & Evaluation
- LLM classifiers, prompt engineering, red-teaming, offline evaluation against human-labeled data, precision/recall tradeoffs, human-in-the-loop and rater systems
- Engineering
- Python, Hack/PHP, Go, Swift, Kotlin, JavaScript, SQL, GraphQL, React, distributed data pipelines, iOS, Android
- Domain
- Trust & safety, content moderation, adversarial abuse, impersonation and identity verification, election integrity
Education
University of Waterloo
Class of 2021Bachelor of Computer Science (Co-op)