FBA AI Science and Analytics accelerates FBA's AI-native transformation by building, integrating, and scaling AI-powered data & science products and seller-facing experiences that drive operational efficiency and growth across Fulfillment by Amazon globally. We're seeking a creative, industrious, customer-obsessed PM who is passionate about applying GenAI techniques to solve real-world data and science challenges at scale.
As a Product Manager Tech in our FBA AI Science and Analytics organization, you will lead the strategy and execution of multiple critical GenAI initiatives, including but not limited to: (1) a Semantic Data Layer that delivers semantically consistent, disambiguated data through a machine-readable interface to every AI application across FBA—eliminating per-application integration costs, metric re-derivation, and the concept-entity ambiguity that drives inaccurate AI outputs; (2) an AI-Powered Anomaly Detection system that autonomously identifies, diagnoses, and surfaces data quality issues and business metric anomalies; and (3) a Data Deep-Dive Agent that autonomously investigates business questions through data mining—surfacing root causes, trend shifts, and actionable insights without manual exploration.
You will drive product innovation across unified, secure, and intelligent AI solutions used by seller-facing application builders across FBA, as well as end-to-end AI-native experiences that solve high-frequency seller workflows — from inventory optimization and inbound efficiency to demand shaping and capacity planning. Your work will directly shape a future where AI is not a separate tool for sellers to adopt, but a native layer woven into every seller's decisions, operations, and growth strategies — delivering actionable insights in minutes rather than days.
Key job responsibilities
• GenAI Product Strategy & Roadmap: Define and execute the product vision for LLM-powered science and data agents, spanning prompt engineering strategies, retrieval-augmented generation (RAG) architectures, semantic parsing, and agent orchestration frameworks that enable autonomous reasoning at scale. Collaborate with scientists to evaluate trade-offs between fine-tuned domain models and in-context learning approaches for optimal accuracy-latency balance.
• Semantic Layer & Knowledge Graph Design: Own the product requirements for FBA's centralized semantic layer, defining ontologies, entity-relationship schemas, disambiguation logic, and machine-readable metadata standards that LLMs consume to produce accurate, governed outputs. Work with scientists to develop techniques that minimize hallucination in downstream agents.
• LLM Agent Development: Partner with applied scientists and engineers to design, evaluate, and ship production LLM agents across multiple FBA domains for internal and seller-facing applications. Define evaluation frameworks incorporating composite accuracy metrics, confidence calibration, chain-of-thought verification, guardrails, and human-in-the-loop fallback mechanisms.
• Stakeholder & Science Partnership: Work closely with the science and tech teams on model selection, fine-tuning strategies, retrieval pipeline optimization, and agent loop engineering. Translate research advances in LLMs, tool-use, and multi-agent systems into shippable product features.
• Metrics & Evaluation: Define success metrics grounded in agent accuracy, query correctness, latency, coverage, and real time-savings. Own end-to-end evaluation including LLM-as-judge frameworks, human evaluation protocols, and A/B experimentation for agent capabilities.
• Rapidly Evolving AI Landscape: Stay current with advances in foundation models, agentic architectures, harness engineering, semantic parsing, and enterprise AI tooling. Identify and prototype emerging capabilities—such as multi-step reasoning, self-improving agents via reflection loops, structured generation with constrained decoding, and tool-augmented inference—for incorporation into the product roadmap.
A day in the life
Your day-to-day responsibilities will include conducting user research, analyzing adoption and productivity data, and making data-driven decisions to guide product development. You'll oversee the entire product lifecycle, from initial concept through launch and iteration, ensuring that products meet quality standards, earn user trust, and deliver measurable business value. You'll partner closely with engineering and science teams to define v1 experiences that prove concepts quickly, measure impact in real time, and stop what isn't working even when it's popular. Additionally, you'll be responsible for developing launch plans, creating product documentation, and establishing processes for measuring and reporting on product performance, adoption metrics, and AI-driven productivity outcomes.