Kubernetes is dramatically complex to the point where I refuse to touch it if I have to do any configuring myself. Unfortunately there are not really any good alternatives. I work on the Azure stack myself, and it has only Azure Container Instances, which Microsoft recommends to not use for production purposes (without explanation why). For deploying a simple machine learning model on Azure Machine Learning, you need to set up a whole Kubernetes cluster, and if I remember correctly it also needs to have a minimum of 6 machines which is a very high amount for simple projects. At least with Azure ML they have an abstraction layer over Kubernetes so that you don't have to deal with any of its intricacies.