Software Development Engineer - Agentic Data Infrastructure, Partner Data And Analytics
Posted Updated
Description
The Data Services team within PDNA (Partner Data And Analytics) builds and operates the Agentic Data Platform that powers Amazon's Selling Partner analytics infrastructure. We serve 900K+ Selling Partners and 30K+ Amazonians through 4,000+ standardized metrics, real-time streaming pipelines validated at Black Friday scale (142M+ events), and the Curated Data Layer (CDL) — Amazon's strategic data infrastructure for Selling Partner GenAI applications including Seller Assistant, Supply Chain Expert, MCF, and ISS. We are building the future where intelligent agents autonomously operate data pipelines, serve analytics through MCP servers, and enable self-service insights at scale.
You will build the infrastructure layer that sits between raw data and intelligent agents. Think of it as the operating system for agentic data — the MCP servers agents call, the query engines they reason through, the schema contracts they interpret, the streaming backbone they react to, and the validation services they trust. This is systems engineering applied to the agentic AI problem: you are not building agents — you are building the infrastructure that makes agents possible.
Key job responsibilities
1) Design and build MCP servers and data access APIs
2) Engineer the real-time streaming backbone using Spark Structured Streaming, Kafka/MSK, Kinesis, and Lambda that powers agentic workflows.
3) Build agent orchestration infrastructure — the tool registries, execution planners, state stores, and handoff protocols that enable specialized agents (ingestion, validation, delivery) to coordinate across the data operations lifecycle.
4) Develop trust and validation infrastructure that keeps agent-generated outputs reliable
5) Extend the lakehouse for agentic consumption — building metadata contracts, schema registries, and catalog APIs on top of our Iceberg-based infrastructure that agents can reason over autonomously.
6) Own your systems end-to-end — design, build, test, deploy, operate, and document. Your infrastructure will be observable, self-healing, and resilient. You will participate in on-call, drive root-cause analysis, and permanently resolve operational issues.
7)
You will build the infrastructure layer that sits between raw data and intelligent agents. Think of it as the operating system for agentic data — the MCP servers agents call, the query engines they reason through, the schema contracts they interpret, the streaming backbone they react to, and the validation services they trust. This is systems engineering applied to the agentic AI problem: you are not building agents — you are building the infrastructure that makes agents possible.
Key job responsibilities
1) Design and build MCP servers and data access APIs
2) Engineer the real-time streaming backbone using Spark Structured Streaming, Kafka/MSK, Kinesis, and Lambda that powers agentic workflows.
3) Build agent orchestration infrastructure — the tool registries, execution planners, state stores, and handoff protocols that enable specialized agents (ingestion, validation, delivery) to coordinate across the data operations lifecycle.
4) Develop trust and validation infrastructure that keeps agent-generated outputs reliable
5) Extend the lakehouse for agentic consumption — building metadata contracts, schema registries, and catalog APIs on top of our Iceberg-based infrastructure that agents can reason over autonomously.
6) Own your systems end-to-end — design, build, test, deploy, operate, and document. Your infrastructure will be observable, self-healing, and resilient. You will participate in on-call, drive root-cause analysis, and permanently resolve operational issues.
7)
Skills
Sign in to read and join the discussion.