KubeCon 2026
The third KubeCon + CloudNativeCon in India, this time in Mumbai. Two days of AI factories, population-scale infrastructure, and a community that keeps getting bigger.
Ten months after Hyderabad, KubeCon + CloudNativeCon India was back for its third edition, June 18 and 19 in Mumbai. The keynote screens were full of population-scale stories: sovereign AI in India's own languages, cricket streamed live to 72 million fans, banking compliance turned into code. And it started with a very good five minutes, when the CNCF handed over the Golden Kubestronaut kit in person.
Mumbai, at full volume
Every edition of this conference gets bigger, and Mumbai was the loudest yet. The venue filled up before the opening keynote and never really emptied, and by evening the screens were pointing everyone to the welcome reception. If Hyderabad proved India could host a KubeCon, Mumbai made it feel routine: same community, more of it.
What I actually went for
The AI thread that started in 2025 has fully taken over: this year it was llm-d for distributed inference, AI gateways in front of production LLMs, agent observability, and GPU debugging war stories. Around it, the things I care about most kept their place: platform engineering, cost as a platform responsibility, policy-as-code, and Kubernetes upgrades that do not ruin your week. One pick per overlapping slot, each linking to its page on the schedule. A personal favourite for the title alone: 'The Death of the YAML-Engineer', which shares a name with an essay on this site.
- Keynote: Sovereign AI at Population Scale. How Sarvam Runs 100 Million+ Conversations Across India's Languages on a CNCF StackAvi Kothari & Vinayak Gavariya, Sarvam
- Keynote: Plug in and Scale: Serving LLM Models on Kubernetes Made SimpleShrinidhi Venkataraman & Nithin R, AstraZeneca
- Keynote: From Platforms to AI Factories. Has Kubernetes Solved It?Saiyam Pathak, vCluster
- Keynote: How JioHotstar Streams Cricket to 72 Million Fans in Real Time with Kubernetes and OpenTelemetryPradeep Bishnoi, JioStar
- What Did My Agent Do? Observability and Accountability for AI AgentsIshan Jain, Grafana Labs
- Beyond VLLM: Distributed LLM Inferencing With Llm-d on KubernetesRavindra Patil, Red Hat
- The Ultimate Kubestronaut Trivia Challenge!Hosted by Fazlur Rahman Khan
- Unity in Diversity: Architecting 'Shared-First' Kubernetes Platforms for Life-Critical WorkloadsManoj K R, Siddiq Tanveer M A, Rishi Nikhilesh Damerla & Geethika Chappidi, Motorola Solutions
- Building and Orchestrating Real Production-ready Agentic AI SystemsKevin Dubois & Daniel Oh, IBM
- Kubernetes Ate the Budget: How We Made Cost a Platform ResponsibilityBhavani Indukuri & Aman Chandna, DigitalOcean
- From API Events To Production in Minutes: Backstage as the Front Door To Multi-Tenant ProvisioningAditya Soni & Atul Bhardwaj, Sailpoint
- Lightning Talk: Secure GitOps for Regulated Workloads: Argo CD Meets Confidential ContainersJitendra Singh, Microsoft
- Building a Self-Service Platform for Stateful AI AgentsShuva Jyoti Kar, Cisco Systems
- Keynote: The Human Cost of Security in the AI EraSonali Srivastava, Improving
- Keynote: Automating RBI Compliance With Unified Policy-as-CodeJim Bugwadia, Nirmata & Ashwath Kumar, Razorpay
- Keynote: One Deploy Plane, Every Runtime: How Shopify Generalized Cloud Native Beyond KubernetesAqeel Nazeer, Shopify
- When LLMs Hit Production: Why You Need an AI GatewayGavrish Prabhu, Nutanix
- When Kubeflow Fights Cilium: Debugging 60% Idle GPUs in KubernetesRamkumar Nagaraj & Bingi Narasimha Karthik, Adobe
- How We Stopped a Crypto-mining Attack in Kubernetes Caused by a Next.js RCEAchanandhi M & Tamil Vanan Karuppannan
- Why We Ditched Kube-proxy: Scaling 10M Daily Browser Sessions With Kubernetes EndpointSlicesRajat Khanna, CommerceIQ
- The Leapfrog Upgrade PlaybookYug Gupta, Walmart Global Tech
- The Death of the YAML-Engineer: Engineering 'Invisible' Platforms With Crossplane and ScoreAbhinav Sharma & Mumshad Mannambeth, KodeKloud
The golden kit
The photo at the top is Nate Waddington of the CNCF handing over the Golden Kubestronaut kit: a yellow beanie, a Thule backpack, and a jacket carrying a patch for every certification in the run. It is a small ceremony, but a meaningful one. The certifications were earned alone at a desk over two years; the kit gets handed to you by the community, in a room full of people who did the same climb.
Old friends at the booths
The hallway track picked up right where Hyderabad left off. I caught up with Mumshad Mannambeth again at the KodeKloud booth, mid book signing for 'AI Meets Kubernetes', and this year he was also on stage co-presenting 'The Death of the YAML-Engineer'. A conference is a very different place the second time around: fewer introductions, more continuations.
After his talk we ended up in a long conversation about my own Crossplane setup. I walked him through how I built a user-friendly platform for engineering at my company on top of Crossplane: developers fill in a handful of fields, the platform owns the thousands of lines of YAML underneath. His talk and my essay share more than a title, and comparing notes with the person who taught half the room was a highlight of the trip.
A real Kubernetes service is thousands of lines of YAML nobody wants to own. So I hid all of it behind twenty fields. Here is what that does for developers, and why the YAML engineer isn't quite dead yet.
Paying it forward
Networking at these events is not only about reaching up to industry experts; it works in the other direction too. I spent a good stretch of the conference with young engineers just starting out, including a graduate intern I met on the floor, swapping the kind of lessons that never make it into documentation. What started as career advice turned into a long conversation about AI and how it is reshaping software engineering as a career. They got insights they could actually use; I got the sharpest questions of the two days.
