Principal Software Engineer - AKS Flex: High Scale

Overview

Imagine working at the forefront of the cloud industry, redefining where Kubernetes can run. The AKS Flex team is building the next generation of Azure Kubernetes Service: a single AKS control plane that can manage worker nodes running anywhere, whether in other Azure regions, on-premises data centers, bare-metal fleets, edge sites, or other clouds.

We are hiring a Principal Software Engineer to scale this vision. You will lead the scaling of etcd and the Kubernetes control plane so a single cluster can reliably manage far larger, more geographically distributed, and more heterogeneous node fleets than ever before. You will also build the platform behind flex nodes for AKS and Unbounded, our open-source project for joining any Linux machine to a Kubernetes cluster, including provisioning, lifecycle, and the Kubernetes-native APIs that make remote machines manageable at scale.

You will solve deeply technical distributed-systems problems in control plane performance, consistency and durability, node provisioning and lifecycle, security, and reliable production operations. Your work will directly unlock capacity for GPU-hungry AI workloads, data-residency-constrained customers, and organizations that want one operating model for all of their compute.

This is a unique opportunity to define the architecture of a new product area from its early days, represent Microsoft in the open-source community, and shape what hyperscale managed Kubernetes looks like without boundaries.

Responsibilities

Technical leadership

• Own the architecture for control plane scalability and the flex nodes and Unbounded platform, and help define the multi-year technical roadmap for AKS Flex in partnership with product and engineering leadership.
• Lead cross-organization design efforts spanning AKS, Azure compute, storage, networking, and identity, driving alignment on interfaces, tradeoffs, and sequencing.
• Make and defend key architectural decisions around correctness, durability, security, scale limits, and failure modes, and establish the design review and quality bar for the area.
• Represent Microsoft in upstream communities (such as Kubernetes SIG Scalability, SIG etcd, SIG API Machinery, and the Unbounded project), driving proposals and building consensus with external maintainers.
• Mentor and grow senior engineers, and raise the technical capability of the broader team.

Control plane and etcd scalability

• Push the limits of cluster scale by improving the performance, reliability, and efficiency of etcd and the Kubernetes API server under large node counts, high object churn, and heavy watch traffic.
• Analyze and optimize control plane bottlenecks such as write amplification, compaction and defragmentation, watch fan-out, list/relist storms, request prioritization, and storage and memory pressure.
• Lead the design and validation of architectural changes to the control plane storage layer, including approaches to partitioning, offloading, and caching state, while preserving correctness and durability guarantees.
• Build scale and performance testing infrastructure, benchmarks, and SLOs that let us measure control plane limits and prevent regressions before they reach customers.
• Build automation for safe etcd operations at fleet scale: backup and restore, member replacement, quorum recovery, upgrades, and repair.

Flex nodes and Unbounded platform

• Design and build the systems that let worker nodes in other Azure regions, on-premises data centers, edge locations, and other clouds securely join an AKS cluster and behave as standard node pools.