Hi everyone,

I’m relatively new to ML and Feast and I’m trying to understand how the aggregation metadata (used to apply transformations on raw data) is connected to the registry.

What I’d like to achieve:
I want to have a single transformation job (or the smallest number possible) that receives configuration telling it:

• how to aggregate the data, and
• in which FeatureView the result should be saved.

This would allow much better reuse and greatly reduce the effort of creating/maintaining multiple transformation jobs.

Am I approaching this the right way? Or am I looking in the wrong place?
How do you guys usually handle this kind of scenario?