Kubeflow is an ML platform like Sagemaker or Databricks that you can self-host in a Kubernetes cluster.
Installing/deploying it is as complicated as it sounds, but we've put together an infrastructure project that lets you '1-click' install it even in tiny environments.
The GH repo (also linked in blog) allows you to start Kubeflow in a codespace or small device using a docker container -- this is both good for trying it out and developing it into your own internal ML platform.
https://github.com/treebeardtech/kubeflow-helm
