Software Development Engineer, Items and Offers Platform, Identity Performance and Defect Detection

The PRISM (Product Identity and Semantic Matching) team is responsible for establishing product identity and attributes, lighting up the product detail page, supply chain, and fulfillment use cases across all marketplaces. The portfolio of services and products we own directly impacts both seller and buyer experiences. Being the world's largest product catalog with billions of products and billions of transactions per day, the scale our services support is massive. For years, we've solved product matching at Amazon scale through a sophisticated combination of human-curated rules and traditional machine learning techniques applied to product data — and now, we're reimagining it all. We are at an inflection point. We are rebuilding its legacy rule-based and classical ML matching algorithms into a new generation of LLM-powered AI systems. This is not a research experiment — it is a production rebuild that will redefine how Amazon understands every product on the planet. As a Software Development Engineer on this team, you will help architect, build, and ship the AI-native matching platform that replaces decades of accumulated logic with intelligent, language-aware, generalizable models running at Amazon scale. You will work on genuinely hard, unsolved problems: serving LLM-based inference over petabyte-scale product data, meeting sub-millisecond latency requirements in the critical path of buyer and seller experiences, and engineering systems capable of serving millions of requests per second. You will solve novel computer science problems at the intersection of large language models, retrieval, distributed systems, and large-scale data processing — problems for which no playbook yet exists. Key job responsibilities We are looking for a customer-focused and highly motivated software engineer to help build the software systems that will improve the quality of Earth's largest e-commerce catalog. This role offers you the opportunity to work with collaborative leaders, work backwards from customers, identify problems, propose innovative solutions, relentlessly raise standards, and have a positive impact on hundreds of millions of Amazon customers and millions of selling partners worldwide. If you are energized by taking generative AI from interesting demo to the production substrate of the world's largest catalog, this role was built for you. About the team The PRISM team advances state-of-the-art GenAI to deeply understand and uniquely identify every product at Amazon scale. We build the canonical product identity and matching system that powers Amazon's universal catalog: eliminating duplicates, preventing abuse, and integrating the world's product data to power agentic shopping and trusted discovery for customers worldwide. We solve foundational challenges including schema-agnostic multimodal understanding, efficient inference at reduced cost, and rapid innovation through automated experimentation powered by agentic systems.