Software Development Engineer, Measurement, Ad Tech, and Data Science (MADS) Foundations- Traffic

About this role

Description
The Traffic Applications team, part of Measurement, Ad Tech, and Data Science (MADS), owns the data ingress system for all of Amazon Ads. Traffic ingests hundreds of billions of ad events daily from upstream systems, decorates them with dimensional and identity data, and publishes structured traffic and billing datasets in a standardized format to downstream consumers across the Ads ecosystem. The system operates under strict SLAs for data freshness and accuracy on tier-one advertising metrics, owning three of the five critical metrics used for Ads billing. As the first MADS system to encounter regulatory and identity changes, Traffic's architectural decisions propagate across the entire MADS stack, making design choices here uniquely consequential.


You will have the opportunity to work on critical business initiatives that directly impact Ads VP and SVP level goals. These include compliance initiatives such as GDPR and ePrivacy, enabling geo-targeting features, and driving architecture scalability to support spiky traffic from live sports events.


Key job responsibilities


As a SDE , You Will

• Own team architecture and lead design on multi-engineer efforts across our Spark-based EMR pipelines, decoration jobs, and publication systems
• Navigate ambiguous technical problems with conflicting constraints — regulatory deadlines, performance requirements, cost targets, and cross-org dependencies
• Identify one-way-door decisions, proactively address architectural deficiencies, and ensure Traffic’s design doesn’t limit what downstream teams can build
• Drive adoption of engineering best practices and maintain sound operations — alarms, telemetry, runbooks — for a system with tier-1 SLAs
• Mentor engineers, contribute to recruiting, and lead constructive technical dialog within the team and across Ads, Customer Trust, and identity-owning upstream systems

What we’re looking for?

• Expertise in large-scale distributed data proce...