Agentic AI & LLM Products · Platform Strategy · Team Leadership
Product leader with 14+ years at the intersection of AI products, data platforms, and go-to-market technology — currently directing engineering teams shipping agentic AI products at Amazon.
I lead a product and engineering team at Amazon that ships the AI products, data platforms, and go-to-market technology powering developer experiences across Amazon's Devices & Services organization — defining the “what” and “why” while directly leading my engineering team on the “how.”
Before this, I spent over a decade in product management, marketing operations, and data strategy roles for some of the world's biggest brands (Nike, Microsoft). I'm passionate about helping my customers and coworkers do their jobs smarter and faster with access to the right tools and insights.
The areas where my experience runs deepest.
Translating business problems into clear product direction. Defining multi-year roadmaps that align engineering investment with measurable business outcomes and operational excellence.
Directing LLM-powered products for real users at scale — from an agentic developer assistant and MCP-based knowledge tools to self-service analytics — and defining measurable quality acceptance criteria as release gates, not aspirations.
Designing shared technology platforms that serve as force multipliers across organizations — the systems that make other teams more effective.
Keeping a variety of tools including content management (marketing web pages and technical documentation), email, analytics, personalization, and developer communication systems secure, compliant, and ready for business users.
Directing analytics products that use Generative AI and traditional machine learning to deliver insights that enable marketers and business owners to fine tune their campaigns and roadmaps.
Leading cross-functional teams of engineers, designers, and analysts. Developing talent, building culture, and managing through ambiguity.
"If the highest aim of a captain were to preserve his ship, he would keep it in port forever."Thomas Aquinas
A selection of projects that represent the kind of problems I find most compelling.
Led a three-org decision to consolidate two competing AI retrieval products for developers into a single customer-facing tool, now generally available to external partners. My team owns the documentation and knowledge layer that powers it — developers get one consistent experience regardless of which AI tool they use, and partner teams build on the layer instead of rebuilding retrieval.
Directed the re-architecture of a RAG-based developer assistant into a multi-agent system (launched June 2026): an orchestrator delegates to specialized sub-agents, and partner teams add capabilities by registering a sub-agent or MCP server rather than rewriting core code. Release-gated on measurable quality acceptance criteria — retrieval precision, actionability, follow-through — defined as gates, not aspirations.
LLM products fail quietly: a data agent can return plausible but wrong numbers. My own end-to-end testing surfaced accuracy failures that engineering's success metrics had not caught, and I paused the launch until schema guardrails and validation closed the gap. Defining quality bars — and holding launches to them — is the discipline I bring to every AI product.
Led strategy for a GenAI Voice of Customer platform unifying unstructured feedback across 20+ channels with AI-driven sentiment analysis and automated triage — including commissioning a documented multi-model evaluation when testing showed the incumbent model's output threatened the product's trustworthiness.
Long-form pieces on product strategy, AI, and the craft of building technology platforms.
AI-assisted side projects that deliver unified intelligence.
A geo-powered marketplace for collectors and dealers to discover card shows, trade nights, and estate sales. Built on geocoding APIs to surface events near any location in real time.
Three-agent pipeline that turns a product problem statement into a research brief, PRFAQ, BRD, and build spec. Scoped and specified by me, built with Claude Code — it tested whether bounded multi-agent orchestration beats single-prompt drafting (it does).
Turns any MCP-capable assistant into a personal chief of staff across email, calendar, and weekly reviews. A working stress-test of MCP for real cross-tool workflows — the personal-scale lab for MCP design decisions I make at work.
For individuals and families — build better money habits with tracking and insight across all your accounts, plus investment analysis and strategy.
For gym rats — understand how your sleep and macros impact your gains. Pulls from Apple Health and diet tracking tools like Cronometer to deliver health insights.
