I have some questions on the role of Feature Stores and why they are necessary.
- Why do we need feature stores when we already have the transformation in the pipeline that will result in what is stored in feature stores? Is it purely to save time that would otherwise be spent on transformation?
- What is the role of feature stores in development and in production?
- When it comes to Data Provenance and Lineage, what should we track? Raw Data, Transformation, Features…or all?
- In a Data Centric AI development, data changes alot. The best way to track these changes are actually the transformation modules in the Feature Stores. How does Feature Stores help?