Your feedback is highly appreciated.
Customer data is a treasure, real time prediction is the Key to that treasure and feature store is the locksmith who can make that key.
Model needs features and serving these features at the lowest latency possible is the most critical part in real time prediction and Feature store serves that purpose.
Another important purpose of Feature store is to eliminate data inconsistency between model training & model serving.
Therefore, Feature Store is a change that Data Scientist should embrace to make real time predictions stronger and it’s a tool that Data Engineers should exploit to solve data inconsistency in Machine Learning.
Data Wrangling + Feature store will keep Data Scientist Focused on Model Optimization and that’s the end goal.
Terminology that should be there on our fingertips while working with feature store
- Feature
- Feature Definition
- Feature Group
- Metadata
- Tags
- Feature Store (Offline / Online)
Flash Cards (Accelerated Learning) on Feature Store
Why AWS Feature Store is Falling short to make an impact ?
AWS has to think beyond their platform and has to promote their feature store for external integrations for wider community adoption.
That’s just my perspective.
-Sumit
