- I understand that K8S is beyond Azure, but can we mentioned somewhere (a line) that AKS also an option to run ML workloads in Azure? - Azure batch and AML can have "low priority" virtual machines, this will help to bring down the cost significantly. (Please double check) - Can we also mention that it is possible for AML pipeline to use Databricks as a step, and also, From databricks (using mlflow) we can publish models to Azure ML?