Tech lead at Meta defining technical strategy for AI data governance, privacy infrastructure, and regulatory compliance at scale. Enabled Meta AI on Ray-Ban Meta and other headline product launches.

AI Data Governance & GenAI Privacy — Tech Lead (2024–Present)

• Designed privacy enforcement architecture for GenAI inference — implemented by a platform team, now processing ~480M data access events/day across every Meta AI product, delivering purpose limitation by default

• Architected automated lineage-based data verification with researcher overrides — replacing heavyweight, high-false-positive approaches with fast bottom-up compliance scans

• Drove data lineage platform adoption across 4+ compliance use cases; proposals convinced partner leadership to maintain alignment through rapid GenAI priority shifts

• Translated EU regulatory requirements (GDPR, data protection) into engineering milestones across policy, legal, product, and ML teams

• Scaled team velocity through technical onboarding, documentation, and mentorship

Privacy Infrastructure & Platform Reliability (2023–2024)

• Built proactive observability (staleness detection, anomaly alerting) shifting from reactive to platform-detected issue discovery — adopted by ads measurement and sensitive data protection teams

• Rewrote privacy policy enforcement system (~2.9B QPS) into modular architecture; implemented automated circuit-breaker with 5-min anomaly shutdown for a critical-path system

• Authored system design RFC, led core deletion implementation and stale data rollout at scale

Earlier: ML infrastructure (2020–22) — built debugging tools reducing cross-worker investigation 4–5hrs→30min, used by hundreds of engineers. Conducted 100+ technical interviews.

• Conducted research in adaptive control for a custom-built variable-geometry aircraft

• Designed/implemented novel control algorithms in C for embedded system, prototyping performed in MATLAB

• Worked with STM32 (ARMCortex-based) MCU

• Developed method for UAVs to retrieve moored or drifter environmental probes from water

• Taught new undergraduate researchers the basics of ROS.

• Technologies and skills used: C, MATLAB, Python, Simulink

• Designed and implemented new lightweight crash-telemetry logger to validate accuracy of existing systems and act as a backup info source if logfiles are corrupted; system now in production and under consideration by both Instagram and FB to become the foundation of future crash logging systems

• Optimized file I/O library to maximize logging performance in iOS; common logging routines now 4% faster

• Designed and implemented oncall tool for reliability engineers automating root-causing of large crash spikes, making it much faster and less mentally taxing to find the bug. Oncall engineers now need less in-depth knowledge of 1) common crash patterns and 2) the tens of tools needed to perform manual root-causing, allowing new engineers to be effective at the role in a week rather than months. Tool is widely adopted by reliability oncall

• Designed and implemented system to automatically determine if changesets attached to a crash task successfully fixed the bug. System is being integrated into the new generation of crash task infrastructure.

• Technologies and skills used:C++, Objective-C, Hacklang, React, Buck, SQL

• Saved an estimated 3 hours of work daily for iOS and Android App Reliability oncalls by automating reliability oncall task management

• Designed and implemented new experiment crash info tool to enable product engineers to much more quickly root-cause bugs in their experiments at-a-glance; tool is now the go-to default for product and reliability engineers across FB for detecting, measuring

severity of, and debugging crashes caused by experiments.

• Created and maintained new user crash-telemetry data pipelines in Hive for automated bug task management, ingesting >1million data points every day

• Worked alongside the Infer static analysis research team to demonstrate information from static analysis tooling research was relevant to critical real-life user crashes; the heuristics-based proof-of-concept system I designed and implemented to match Infer info to stack traces provided repro assistance for 6% of daily top bugs

• Technologies and skills used: Python, Hive, Presto, SQL, Hacklang, React, large-scale data processing

• Modernized critical internal virtual machine tools hosted on a synchronous webpage and used by Tier 3 Product Support and product development employees around the world to diagnose and develop software on different platforms.

• Continued development of a new API layer over VMWare's VCloudDirector, automating commonly performed tasks, and contributed to a new asynchronous frontend design and implementation in AngularJS.