Nikhil Kaul is a product manager on Microsoft's Azure Kubernetes Service (AKS) team in Seattle. He works on cost and efficiency: getting more useful work out of Microsoft's 200M+ core fleet, and building the tools that let AI agents run Kubernetes clusters well.

Capacity & efficiency

Compute is the constraint on everything in AI right now, and a lot of it is used badly. Nikhil's job is to fix that inside Microsoft's own fleet. He built the measurement model the org uses to tell whether efficiency work actually paid off, and he sets product direction across Azure Kubernetes Service and Virtual Machine Scale Sets on the levers that move the number: right-sizing workloads, mixing SKUs so placement is cheaper, and Karpenter-based node auto-provisioning.

AKS-Skills

In 2026 Nikhil led the public launch of Azure/AKS-Skills, an open-source set of seven skills that teach AI coding agents how to operate AKS: troubleshooting live incidents, looking up known issues, cutting costs, running GPU and inference workloads, capturing network traffic, checking AKS Automatic readiness, and designing clusters. They follow the open Agent Skills standard, so the same set works in GitHub Copilot, Claude Code, Azure SRE Agent, and anything else that speaks it. The skills investigate and ask before they change anything. MIT licensed.

Azure Chaos Studio

From 2023 to 2025 Nikhil was the product manager for Azure Chaos Studio, Microsoft's chaos engineering service. He took over a product that needed a turnaround and grew it to 200K+ fault-injection experiments a year, with some of the world's largest banks running it in production. Then he led a rewrite of the whole platform. It shipped in nine months, cut serving costs 90%, grew capacity 100×, and saves $3M+ a year. He demoed it on stage at Microsoft Ignite 2023 and Build 2024. Along the way he pushed the team toward agent-assisted development, and feature throughput went up roughly 30× in a year.

Earlier: Windows

Nikhil joined Microsoft in 2021 on the Windows team, working on recovery and efficiency for Windows 10. He owned Windows Optional Features, which 20M+ people use every month, and the work that trims Windows down to a smaller install. He also got Wi-Fi drivers into UEFI firmware across four silicon vendors, so a PC can recover its operating system over the network before Windows even boots. That now ships on major OEM devices. And he led the effort to get HP, ASUS, Acer, and Lenovo building Windows-on-Arm devices with the same imaging pipelines they already used for x64.

Recognition & projects

Nikhil holds a granted US patent for Automated Resiliency Validation. The idea came from his Chaos Studio years: instead of injecting a fault and assuming the system held, check automatically that it did. Microsoft recognized it with a Patent Award in January 2026. He also won first place at the Microsoft Global Hackathon 2025, in the executive challenge judged under Azure CEO Scott Guthrie, with an MCP-based scanner that finds resilience problems in AI-generated apps. That project inspired the Azure Copilot Resiliency Agent, now in preview. Photos of both are on the awards page. On the side he has built Agency Co-Work, a desktop workspace for AI agents that ships with its own source code; a top-20 iOS health app that Doctors Without Borders teams use for pediatric dosing; and a Mac mini home lab that runs, among other things, the AI behind this website.

Education

Nikhil has a B.S. from Duke University, where he double majored in computer science and philosophy and spent a semester at Oxford studying philosophy. He was born in Austin, Texas, and grew up moving between the US and India: Texas, North Carolina, Bangalore, then New Delhi, where he finished the International Baccalaureate at the American Embassy School before heading back to North Carolina for Duke. He lives in Seattle.

Contact

Email nkaul2345@gmail.com, find him on LinkedIn, read his takes on X (@TheTakesSmith), look at code on GitHub, or ask his AI. It runs on his own hardware and knows everything on this page.