Your feature, training, and batch inference pipelines can be written and scheduled as Modal Python programs that write/read to/from the Hopsworks Feature Store.
Your Modal serverless Python programs are the compute for your pipelines, while Hopsworks provides state (features, labels, models).