AWS DevOps Agent is becoming a self-correcting operations platform. It measures its own investigation quality, detects when a customer's configuration is degrading, corrects what it can on its own, and tells the customer clearly when it cannot. Today the platform reasons over live topology, learns from customer feedback, and runs release-readiness reviews on production pull requests. About 9,900 internal and 5,900 external accounts depend on it, and that number is growing double digits week over week. We are looking for a Senior SDE to help lead this next phase, turning a fast-growing, composable agent platform into one that provably gets better the more it is used. You will work across three core teams (System of Context, the Investigation Agent, and Evaluation) on problems that do not yet have a settled technology strategy. How should an agent decide a piece of context is stale? How should fast and slow reasoning run concurrently and correct each other? How do we prove, to ourselves and to the customer, that quality is real and improving rather than just claimed? This is a role for someone who wants to be hands-on with hard agent-architecture problems (memory, evaluation, automated reasoning, multi-agent coordination) while also shaping the technical strategy that AWS service teams and enterprise customers will depend on for incident investigation and site reliability. Key job responsibilities You will lead the design and delivery of platform-level capabilities spanning multiple teams, including event-driven context freshness, concurrent fast and slow investigation, and real-time evaluation checks. You will take on problems where the customer need is understood but the right technical approach is not, owning technical decisions with system-wide consequences around memory lifecycle, retrieval strategy, model-aware routing, and the boundary between what an agent should self-correct versus escalate. You will drive design reviews across System of Context, Investigation Agent, and Evaluation, making sure the three converge on one coherent architecture. You will stay hands-on, personally driving the highest-leverage code and code reviews, including the prototype that sets direction, the interface that defines a team boundary, and the fix that restores operational stability. You will mentor and raise the bar for engineers across the three teams, and partner with Product, Applied Science, and adjacent AWS teams to turn research and strategy into shipped, measurable improvements. A day in the life You might start in a design review for how Production Intelligence decides topology is stale, weighing automated reasoning against simpler heuristics. Mid-morning, you pair with an engineer on real-time reasoning checks. After lunch, you review a memory-consolidation design, pushing back on claims not yet backed by evidence. Later, you are deep in code on the interface between investigation agents or the recall-logging hook that tracks memory confidence. You close the day documenting a search-versus-knowledge-graph decision. Other days are more focused, spending a week on one hard implementation or running evaluation experiments. About the team AWS DevOps Agent is a platform, not a single agent. Its foundation is built by three teams working as one system. System of Context owns Production Intelligence, memory, and the context pipeline. Investigation Agent owns the execution engine and the fast and slow reasoning architecture. Evaluation owns the platform's immune system, the checks and scoring that keep quality honest. We work in a tight loop where context is supplied, tested against reality, and scored for improvement. This role works across all three. We treat context quality as the real differentiator and hold ourselves to a high bar on evidence.