Description

We are looking for a Software Development Engineer to join the EKS KCP Scalability team and work on some of the hardest distributed systems problems at Amazon. You will design, build, and operate systems that directly determine whether EKS customers — from startups to the largest AI/ML workloads on the planet — experience a reliable, performant control plane.

This is not a role where you implement features in isolation. You will work across the full stack: from the Kubernetes API server process and upstream community engagement, through autoscaling services that right-size control planes in real time, to the SLA measurement pipelines that hold us accountable to our customers. You will own systems end-to-end — from design through production operations — and your work will be measured by customer outcomes, not lines of code.

Key job responsibilities
You will build and operate the Vertical Auto-Scaling Service (VAS) and its next-generation successor (VAS 2.0), which dynamically right-sizes EKS control planes by evaluating CPU/memory utilization, etcd throttle rates, node-count thresholds, and network utilization simultaneously. You will work on the SLA measurement pipeline (MinutelySLA → DailySLA → MonthlySLA) that enforces EKS's uptime commitments, investigating breaching clusters weekly and building automation to detect and mitigate degradation before customers notice.

You will contribute to the control plane architecture for EKS Ultraclusters, defining how the API server, etcd, and associated components scale to support 100,000-node clusters running generative AI workloads. You will maintain and extend version release qualification scale tests that gate every new Kubernetes version before it reaches customers. You will engage with the upstream Kubernetes community — driving KEPs that work backwards from EKS customer requirements around performance, scale, and resiliency.

Depending on your interests and the team's priorities, you may also work on workload identity systems (IRSA, EKS Pod Identity), Cluster Access Management, EC2 capacity management and grey failure detection, or Large-Scale Event response and weight shifting.

About the team
The EKS KCP Scalability organization owns the performance, availability, and autoscaling of the Kubernetes control plane powering Amazon EKS — from small development clusters to 100,000-node Ultraclusters running generative AI workloads. We ensure every EKS cluster operates within its contracted SLA and delivers predictable, high-performance behavior at any scale.

Our charter spans three domains: Performance , Availability + Autoscaling , and Auth. We operate at the intersection of distributed systems, Kubernetes internals, and AWS infrastructure — building systems that scale to hundreds of thousands of clusters globally.