Technical leader for feature engineering and ML-powered risk decisioning systems supporting abuse prevention across Amazon’s global customer base.
Key Accomplishments
- $800MM/year impact: Led architecture and engineering strategy for Buyer Abuse Prevention across e-commerce domains; designed distributed, multi-tenant ML feature platform processing 10k+ TPS
- Account clustering migration: Led migration of key risk signal with 500MM/year impact; architected phased rollout and shadow testing framework
- Performance-accuracy trade-offs: Orchestrated 50+ detection points spanning returns, checkout, and payments; balanced sub-150ms real-time decisions with comprehensive offline evaluations
- Customer service partnership: Drove strategic alignment delivering $1.5MM/year savings while designing enhanced experience paths for high-risk interactions
- Engineering-ML bridge: Created feature engineering guides and cross-functional workshops, reducing support queries by 22% and eliminating integration bottlenecks
- Scaled mentorship: Conducted weekly office hours, authored best practice guides, and fostered feedback culture across 6 global teams (~60 engineers)
- 33% infrastructure cost optimization: Led hands-on workshops for JVM profiling, instance right-sizing, buffer reduction, and dynamic scaling
- AI adoption: Used Claude 3.5 LLM to build migration accelerator for 250+ workflows across 10 teams; created strategic roadmap for broader adoption
Technologies: Java, TypeScript (AWS CDK), Python ML pipelines