At launch, PS can be installed in the cloud, but it won't be able to leverage an existing Databricks or Sagemaker environment at launch. Instead you would just install on directly on one or more compute instance (e.g AWS EC2).
That being said, PS architectured to support other execution environments, so the challenge here isn't technical but rather defining the requirements for the integration. Do you have any more information on what you'd want out of an integration with Databricks or Sagemaker that can help us prioritize those integrations?
We are spinning up a collaboration effort with AWS under their Machine Learning Industry Solution, MLIS, program. While support for Sagemaker isnt a requirement it would help win their sales peoples' commitment and invitation to more accounts. EC2 is a good start, but I know they will raise the question.
Wrt Databricks, they've mostly disengaged and did let us know that until we support Delta Lake and execution of models (I believe our Spark tool fell behind) there isnt much for us to do together. If/when we can go back with a compelling collaboration story we are ready to open the door again.
Thanks Hakan for the insight! There are certainly some interesting possibilities with respect to a Sagemaker integration. While not currently priorities, two specifically come to mind: