I have a requirement to build a self service machine learning platform on Kubernetes. This platform is to onboard data science teams irrespective of their use cases whether its ML, DL or LLM. I would like to understand how do I go from a Kubernetes design to components like kubeflow, mlflow, spark, kafka etc to even GPU/CPU baselining. Any leads or reads on this matter is highly appreciated. The kubernetes will be an on prem installation and not a cloud managed service.