Software Development Engineer III, Omnichannel Metrics (OCM) - Measurement, Ad Tech, and Data Science (MADS)

Every time an advertiser runs a campaign on Amazon, they want to know: did it drive sales at Target, Walmart, Costco, and hundreds of other retailers? Omnichannel Metrics (OCM) answers that question — and we're just getting started. We build the ML-powered attribution systems that connect billions of ad impressions to real-world purchases across the entire retail landscape, processing massive-scale panel data to give advertisers measurement insights no competitor can match. OCM is one of Amazon Ads' highest-growth products, directly responsible for unlocking advertiser investment by proving the full omnichannel value of Amazon media. Right now, we're shipping some of the most ambitious projects in MADS: Fine-Grained OCM that breaks attribution down to category and retailer level for the first time, OCM for Autos bringing omnichannel measurement to the automotive vertical, OCM Local building location-based foot traffic measurement for restaurants and retail, a Combined Methodology that unifies our Panel and Circana measurement into a single omnichannel view, and international expansion of the Unified Panel to 8 new marketplaces. Looking ahead to 2027, we're building real-time streaming attribution, integrating LLMs for agentic report generation and automated campaign optimization, expanding into healthcare and home services verticals, and launching causal incrementality measurement. The problems are genuinely hard — privacy-safe identity resolution at scale, calibrating ML models across sparse panel data, building streaming systems that produce daily estimates from weekly ground truth — and the ambiguity is high. You won't be handed a spec; you'll be defining what to build and how. As an SDE III, you will own the architecture and delivery of major system components end-to-end. You'll make technology choices that shape how OCM scales from thousands to tens of thousands of active studies, design distributed data pipelines processing billions of events in Spark/EMR, and build the AI-powered features that transform OCM from a reporting tool into an autonomous measurement platform. You'll work across Scala, Java, and Python, with deep expertise in big data technologies (Apache Spark, Kestrel, EMR), AWS infrastructure (Lambda, DynamoDB, Step Functions, CDK), and data modeling for both relational and non-relational systems. You will lead technical design reviews, mentor engineers, and raise the engineering bar across the team. You'll partner directly with applied scientists on ML model productionization and with product managers to translate ambiguous customer problems into scalable technical solutions. This is a team where an L6 engineer has outsized impact — we're small enough that your architectural decisions become the system, and high-visibility enough (S-team and VP-level goals) that leadership sees the results. The ideal candidate thrives in ambiguity, has a track record of delivering complex distributed systems in fast-moving environments, and gets energized by problems that sit at the intersection of big data, ML, and product. You're not just looking for a job — you're looking for a team where you can shape the technical direction of a product that's redefining how the advertising industry measures success. Key job responsibilities - Design and build scalable data pipelines processing billions of ad impression and panel receipt events using Spark/EMR on Kestrel, powering omnichannel attribution for thousands of active advertiser studies - Lead architecture decisions for OCM's calibration, report generation, and data processing systems spanning multiple pipelines across multiple AWS accounts - Own end-to-end delivery of major product features — from technical design through production launch — including real-time streaming attribution, AI-powered report generation, and new vertical measurement frameworks - Drive operational excellence across OCM's distributed systems: improve observability, eliminate silent