We are seeking a Principal Product Manager - Technical to own answer quality for Alexa for Shopping (formerly Rufus), the conversational shopping AI used by hundreds of millions of customers every month. You will define what a good answer is, build the measurement that proves it, and drive the work that makes it better: deciding what quality even means when there is no single right answer to "what should I get my dad", owning the evaluations that stand in for hundreds of millions of customers, and identifying the opportunities that will make this the best shopping AI in the world before they are obvious to anyone else. The bar you set is what earns that title, and it moves higher every sprint, so you will clear it and then raise it again.
Key job responsibilities
You will own answer quality strategy and roadmap, driving clarity where the problem and opportunity are not yet defined. You will own the quality metric itself and the evaluation system behind it, including the agentic judges, the benchmark missions, and the competitive comparisons that tell us whether we are ahead, and you will invent the successor metric when today's stops separating good from great. Your charter is not to report the number but to move it, finding where the largest quality gains are available, turning them into work for the teams who can deliver them, and making the call when a change that lifts one dimension costs another. You will partner closely with multiple engineering and science teams and will drive the cross-organizational decisions to define and influence the future of shopping AI.
About the team
Our mission is to fundamentally change how customers shop on Amazon by leveraging the state of the art in AI. Every month, hundreds of millions of customers use Alexa for Shopping to help with shopping decisions. We are building a conversational shopping experience that understands what customers need, provides trustworthy and helpful responses, and makes the entire shopping journey effortless. We will build the best shopping AI in the world — and then do it again in the next sprint.
