About
Most AI leaders can tell you what they built. The harder question at scale is what to decide, and how to build an organization that can execute it under real ambiguity. That's the work I do: I set the multi-year technical strategy for how 500M+ customers experience personalization and applied AI at Amazon, then build the teams that turn strategy into shipped products.
In 2025 I was the single-threaded owner of Help Me Decide, a GenAI shopping experience spanning Amazon's highest-traffic surfaces. I made the deliberate confidence-over-coverage bet, trading catalog breadth for customer trust, and drove the architecture and cross-org execution behind it. It shipped as Amazon's highest-performing weblab of 2025 ($451M in Order Product Sales) and was cited in the company's Q3 earnings.
I lead a 75+ person org of principal engineers, applied scientists, and engineering managers, and my influence extends across 150+ people in Search, Ads, and Home. I raise the hiring bar, develop managers-of-managers, and set the conditions where ambitious AI actually ships. Sixteen-plus years in, the through-line is simple: I build AI that earns trust at scale, and the orgs that sustain it.
Experience
- Single-threaded owner of Help Me Decide. Set the technical strategy and made the confidence-over-coverage architecture bet (trading catalog breadth for trust); shipped as Amazon's #1 weblab of 2025 ($451M OPS) and cited in Q3 earnings.
- Own the multi-year vision and investment strategy for Amazon's recommendation and customer-intent platforms, cascading it into roadmaps across a 75+ person org and 150+ people in Search, Ads, and Home, tied to $12B+ in attributed revenue (2025, +120% YoY).
- Build the organization: lead principal engineers, applied scientists, and managers-of-managers, raising the hiring bar, scaling the org, and developing leaders so teams execute autonomously at a high bar.
- Chose a workload-specific model strategy for real-time explanation generation (Sonnet for reasoning quality, fine-tuned Mistral for control), accepting operational complexity in exchange for quality and compliance headroom.
- Own the real-time customer-intent platform at 500K RPS / 45B daily transactions, powering personalization across Search, Homepage, Cart, and Detail Pages.
- Drove adoption of an internal AI agent platform (persistent memory, multi-agent orchestration, secure Slack/Quip/M365/AWS integrations), scaling it from Personalization to multi-org use to accelerate engineering productivity.
- Set direction for the next-gen recommendation stack (LLM-generated customer profiles, embeddings, retrieval, and transformer rankers), powering explainable shopping guidance across major surfaces.
- Transitioned from Principal Engineer to Engineering Manager, leading 15 global engineers.
- Led implementation of a recommender engine on Google Vertex AI with custom embeddings, plus a multi-tenant real-time personalization engine with multi-language and A/B testing support.
- Re-architected to event-driven microservices on Kubernetes for HA/failover, cutting downtime 40%, and adopted gRPC to reduce request processing time 50%.
- Built distributed pricing, personetics, and deposit engines with real-time alerting for the Digital Wealth team.
- Led backend for an AI-powered student web-filtering system serving 10M+ students globally.
- Introduced real-time content classification (NLP + ensemble URL scoring), cutting review delays 70%.
- Led engineering and technical direction as an early-stage CTO, owning architecture and the founding build.
- Co-founded the company and built the product and engineering org from zero.
Technical Skills
◆ Languages
✦ ML / AI
▲ Infrastructure
● Databases
★ Leadership
Notable Projects
Help Me Decide (HMD): flagship AI launch
Single-threaded owner of Amazon's GenAI shopping decision feature. Set the strategy, drove the confidence-over-coverage architecture, and orchestrated engineering, science, and product across my org. $451M OPS in a three-week A/B test, the #1 Amazon weblab of 2025, cited in Q3 earnings.
Read the technical deep dive →Cross-Amazon Personalization Guild
Co-founded an internal guild that created shared libraries, best practices, and hiring rubrics across Amazon AI teams, scaling engineering quality and the bar for talent beyond my own org.
Real-Time Behavioral Personalization Layer
Built a near-real-time streaming pipeline (Kinesis + Flink + DynamoDB) that enabled session-aware ranking across Amazon.
Go deeper →
The leadership decisions, architecture, and trade-offs behind Help Me Decide, the $451M launch I owned end-to-end, written up in detail.