Meta

Workforce Management Products

• Led an end-to-end rebuild of Meta’s scenario planning application, used to model and execute company-wide reorgs, from an in-house stack to Next.js/TypeScript, enabling 4× faster deploys and shortening feedback-to-ship cycle from weeks to days.
• Designed and shipped 14 tool-calling GraphQL APIs powering an in-product AI agent that drives multi-step workflows, from drafting a plan to approval and execution, with async operations and per-user permission checks.
• Scaled reorg plan processing throughput 3× and cut p95 latency by 40% by building automated load tests, profiling bottlenecks under stress, and optimizing production hot paths.
• Cut approval chain generation failures from 30% to <1% by tracing intermittent failures to concurrent jobs reading stale data from read replicas, then enforcing read-after-write consistency, removing re-runs and escalations that delayed plan execution.
• Extended backend services and data ingestion pipelines to onboard 1,000+ retail store workers, replacing a fragmented manual process and cutting hire-to-onboard time by 50%, helping the wearables business expand into new markets faster.
• Built two multi-step AI agent workflows that generate and self-correct unit and E2E tests, enforcing quality through reviewer sub-agents, automated quality gates, and lint rules, lifting team test coverage from 58% to 85% in 4 months.
• Architected and executed the migration of 30 business-critical cron jobs to a serverless platform, cutting failed runs by 60% and keeping 200,000+ records in sync with upstream Workday data for accurate headcount and budget planning.

Workforce Management Products

• Automated hold-or-backfill decisions for vacated roles across 80,000 positions, replacing manual review with an event-driven rules engine in Hack/PHP and GraphQL supporting batch and streaming ingestion.
• Optimized tag transactions to achieve an 80% latency reduction in tag processing speeds by streamlining tagging logic and leveraging a distributed scheduler for parallel job execution.
• Re-designed and optimized position-tagging transactions around a distributed scheduler with parallel job execution, cutting end-to-end latency by 80%
• Built an event-driven pipeline that detects changes via triggers and propagates incremental updates across 5.6M+ records, keeping systems in sync and preventing data drift.