Kubeflow

Kubernetes model serving has changed significantly since 2024. Yuan Tang's article «AI/ML Innovation in the Kubernetes Ecosystem» captured an important transition point. Since then, OCI image volumes have become stable, inference-aware routing has moved into Gateway API, KServe has introduced an LLM-specific serving API, and projects such as LeaderWorkerSet and llm-d have made multi-node and disaggregated inference more explicit. I wrote an update on what has matured, what remains early, and where the operational complexity has moved rather than disappeared. https://lnkd.in/efMTtS4D #Kubernetes #LLMInference #AIInfrastructure #DistributedSystems

🎙️ New Kubeflow Podcast Episode is up! What does it take to bring Arm support to Kubeflow? Why are CPU-based AI workloads still critical in modern ML platforms? And why does Kubeflow continue to play such an important role in the Cloud Native AI ecosystem? In this podcast, Chase C. talks about: 🔹 The journey of bringing Arm support to Kubeflow 🔹 How Ampere contributed to the ecosystem 🔹 The AI and ML workloads that are best suited for CPU architectures 🔹 Why Kubeflow exists and the problems it solves for organizations building AI platforms on Kubernetes 🔹 The future of cloud-native AI and the opportunities ahead Whether you’re a Kubeflow contributor, platform engineer, MLOps practitioner, or simply curious about the infrastructure powering modern AI, this episode offers valuable insights from someone actively contributing to the project. 🎧 Watch the full episode here: 🔗 https://lnkd.in/gQgmDvhz We’d love to hear your thoughts. #Kubeflow #CloudNativeAI #Kubernetes #CNCF #MLOps #OpenSource #Arm #Ampere #AIPlatforms #Podcast

📢 📢 📢 Great news! We have extended the deadline for the Contributor of the Month July nominations until Friday, July 24th. Don't miss the chance to recognize contributors. Submit your nominations today! 🚀 https://lnkd.in/gf9ahK8B

Operating AI/ML Workloads on Kubernetes: A Headlamp Plugin for Kubeflow - https://lnkd.in/gkTuFfJ3 #Kubernetes

The Kubeflow community is cooking! 🧑🍳 As the community continues to grow and work its way towards graduation, we are ensuring we give credit where credit is due to all the maintainers (new and advanced) that help Make Kubeflow.. Kubeflow! Before we dive into that, let's explain why Kubeflow is special. 🧼 📦 Now in my humble opinion Kubeflow is a curated ecosystem of tools built by a community with a mission rooted in three core pillars: 🔓 Sovereignty & Democratization: Giving builders total control over their models, data, and infrastructure. Run it on-prem, in a VPC, or across public clouds—completely free from vendor lock-in. Save yourself from the painful tokenomics. 🚀 Real-World Impact & Value: We don’t engineer in a vacuum. We ended up building for massive, distributed scale, but that started by simply focusing on day-to-day value for the engineering and data science teams relying on our ecosystem. No hype. Just humans. 🤝 Humility & Choice: No single project solves everything, so we meet teams exactly where they are. Deploy the full Kubeflow Community Distribution, or just pull the specific components you need for your custom stack. 🦸 🦸♂️ 🦸♀️ The above didn't happen by accident, the real power is not just the tools (although they are amazing), but the paid and unpaid community members curating and creating our Kubeflow culture! 🥇Tying this back to where we started, Kubeflow now has a way to nominate one of our amazing contributors that helped you in any way. We are looking for submissions and would greatly appreciate if you did that here (https://lnkd.in/gCcUQ6Rr) 🎓 Now, if you are a fan of Kubeflow in general and interested in our graduation efforts, feel free to show some love on this PR: https://lnkd.in/gUM3x-5y. Any support and/or feedback you can provide will help us continue to build and evolve Kubeflow in a way that is valuable to those who depend on it! 🌍 Let's go! 🧗♀️ 🧗♂️ ⛰️

The Kubeflow Graduation Due Diligence document is now open for public comment! This is an exciting milestone in Kubeflow's graduation journey within the Cloud Native Computing Foundation (CNCF). We've reached an important point where all adopter interviews have been completed, all blocking items have been resolved, and the community continues to demonstrate a healthy, active, and collaborative OSS environment. We'd love your feedback! If you're interested in Kubeflow's graduation, please review the application and share your support and comments before July 22. A huge shout-out to everyone across the Kubeflow and CNCF communities who contributed their time, expertise, and feedback to get us here. Special thanks to Faseela K and Brandt Keller for sponsoring our application and helping drive this effort forward. I'm super excited about the future and the innovations we'll build together to advance Cloud Native AI infrastructure. https://lnkd.in/eQWp9wZ9

This is an amazing new initiative from the Kubeflow Outreach Committee to encourage and celebrate contributions across the community 🎉 Don't miss the opportunity to be recognized through the Kubeflow Contributor of the Month program for your valuable contributions and efforts. Every contribution makes a difference!

Celebrating the hands and minds behind Kubeflow! 🚀 We are incredibly excited to announce the launch of the Kubeflow Contributor of the Month program! Kubeflow thrives because of our amazing community. This new initiative is all about showcasing the impact of our contributors, celebrating your incredible achievements, and continuing to grow our open-source ecosystem together. 🗓️ Key Details: The program officially kicks off on July 1st, 2026. Know someone who has gone above and beyond? Sponsors can now submit nominations. Check out the eligibility criteria, learn more about the program, and get ready to celebrate our community. 🔗 https://lnkd.in/gf9ahK8B #Kubeflow #OpenSource #AI #Community #CloudNative

A few months ago, we asked how you're building and managing AI workloads – and the response was fantastic. Thanks everyone who took the time to share your experiences! Today, we're excited to publish the Kubeflow SDK User Survey 2026 results and show how your feedback is already shaping our roadmap. Here are some of the biggest takeaways from 30+ ML engineers, platform engineers, and AI practitioners: - Managing distributed training infrastructure and GPU resources is still too complex - Debugging Kubernetes-based AI workloads remains one of the biggest pain points - The build → push → run workflow slows down experimentation and developer productivity - Modern AI workloads are becoming increasingly sophisticated, requiring better tooling and integrations Based on your feedback, we're already working on several improvements, including: ✅ Make scalable distributed training effortless ✅ Better observability with OpenTelemetry integration ✅ Native MLflow integration for experiment tracking ✅ AI-assisted debugging capabilities ✅ Expanded support for modern AI workloads across the Kubeflow ecosystem This survey reinforces that improving the AI developer experience is about much more than adding features, it's about making it easier to build, debug, and scale AI applications. Read the full survey results and roadmap here: https://lnkd.in/e4F-r5ez Special thanks to Akash Jaiswal for putting together the blog post!

As we announced previously, four (4) vendor seats are open for one-year term, appointed by the Kubeflow Steering Committee (KSC). We've had good early interest, and we want to make sure no one misses the window – so we're setting a deadline. 📅 Please submit your interest by Friday, July 17, 2026. If your organization has a meaningful stake in Kubeflow distributions and the broader ecosystem – as a distributor of Kubeflow subprojects, a contributor to them, or an organization operating Kubeflow at scale – we'd love to hear from you. Please use this survey form to apply: 👉https://lnkd.in/eTPy3sDg If you have any questions, reply in the kubeflow-community-distribution Slack channel or DM any KSC member.