Skip to content

Amazon

NewBe an early applicant

Software Development Engineer, AWS DevOps Agent

Posted Updated

Description

Build the AI agent that investigates production incidents for AWS customers. The AWS DevOps Agent reads the same telemetry a human operator would — logs, metrics, traces, topology — reasons about what went wrong, and delivers a written root-cause analysis in minutes. We're hiring an SDE II to work on the investigation agent: the core reasoning engine that turns an alarm into an answer.

This is not wrapper code over an LLM. You'll work on agent orchestration, tool design over MCP (Model Context Protocol), prompt engineering at scale, sandbox isolation for untrusted code execution, and evaluation systems that measure whether the agent actually got the diagnosis right. The system spans Rust, Python, and TypeScript services across a hundred packages, and ships to production across all commercial AWS regions.

Key job responsibilities
- Design and build agent reasoning systems that investigate production incidents end-to-end — from alarm to root-cause analysis — using large language models and a tool-calling architecture over MCP.
- Own and evolve the investigation agent's tool surface: the capabilities the agent uses to read CloudWatch, X-Ray, Kubernetes, CloudTrail, and service topology on a customer's behalf.
- Build and improve evaluation systems (offline scenarios with ground-truth grading, online trajectory evaluation of production runs) that measure agent quality and catch regressions before customers do.
- Work across the stack — Python agent runtime on Strands, Rust Data Plane services, TypeScript control plane and CDK infrastructure — in a system where a single investigation crosses half a dozen services.
- Design prompt and skill architectures that shape how the agent reasons without hard-coding investigation logic, balancing model flexibility with reliability.
- Collaborate with security, AppSec, and partner service teams on the trust boundaries that let an AI agent safely read customer telemetry at scale.

A day in the life
You might start the morning reviewing an eval regression — the agent's accuracy on a new scenario class dropped, and you're digging into the journal trace to understand what the model missed. After lunch you're pairing with a teammate on a new MCP tool that gives the agent access to deployment history, deciding what information to surface and what to withhold. By end of day you're reviewing a CR that changes how the agent handles multi-signal incidents where logs, metrics, and traces tell conflicting stories. The problems are ambiguous, the feedback loop is fast (run the agent, read the trace, grade the result), and the work ships to every AWS commercial region.

About the team
The AWS DevOps Agent team builds an AI agent that investigates incidents, reviews releases, and answers operational questions about running AWS environments. We own the full vertical: the agent runtime, the tool servers, the sandbox, the evaluation pipeline, and the data plane that records what every run did. Our system serves AWS customers across all commercial regions, and the investigation agent is the flagship — the thing that turns "an alarm fired" into "here's what happened and what to do about it." We're a team that cares about whether the agent is actually right, not just whether it sounds right, and we've built the eval infrastructure to tell the difference.

Skills

Apply

Apply to Software Development Engineer, AWS DevOps Agent at amazon

Create a free account first, or go straight to the employer.

Free RoleWatch account

  • Get an email when new jobs match your search
  • Save jobs and track every application
  • See how well your CV matches each job

Already have one?

No account needed. Opens the employer's own site in a new tab.

See also

DevOps jobs by country — openings, pay and top skills →