End-to-End ARM64 Support & Validation on Kubeflow

ARM64 is becoming increasingly common with the rise of Apple Silicon and cloud instances like AWS Graviton, but Kubeflow still doesn’t run consistently on ARM-based systems. Right now, different components behave differently — some work partially, some require manual fixes, and others don’t work at all. This makes it difficult to use Kubeflow reliably outside of traditional x86 environments . The goal of this project is to make ARM64 a properly supported architecture across the Kubeflow ecosystem. Instead of partial or experimental support, the idea is to make sure Kubeflow works the same way on ARM as it does on AMD64. To achieve this, I plan to start by auditing the container images used in Kubeflow and identifying gaps in ARM64 support. Based on that, I will enable multi-architecture builds using Docker Buildx and integrate them into existing CI/CD pipelines so that ARM64 images are built and published automatically. I will also update manifests to remove architecture-specific references and ensure they work across both architectures without requiring changes. A major part of the work will be validation. I will deploy Kubeflow on ARM-based environments (OCI Ampere) and run end-to-end tests to catch issues that only appear at runtime. Any failures related to dependencies, builds, or configuration will be debugged and fixed incrementally. By the end of the project, Kubeflow should be able to run end-to-end on ARM systems without manual workarounds. The final output will include working multi-architecture builds, improved CI/CD pipelines, updated manifests, and a clear “golden setup” guide for running Kubeflow on ARM. Overall, this project is about making Kubeflow more consistent and usable across modern infrastructure, especially as ARM adoption continues to grow.