The Logical Feature Store: Managing Data for ML
Georgia will introduce the logical feature store, its data management capabilities, and organizational design patterns to support them.
Georgia will introduce the logical feature store, its data management capabilities, and organizational design patterns to support them.
Productionizing machine learning (ML) models remains a significant challenge for enterprises attempting to scale-up ML operations. As a result, many organizations can benefit from a feature store to accelerate time to production and promote feature reusability, reproducibility and reliability. However, not all organizations have the scale and ML use cases to justify implementing a fully-fledged feature store solution, or the resources to do so. Furthermore, when it comes to feature store product selection, variability across the industry and overlap with other technologies causes confusion amongst buyers. Therefore, this talk will introduce the concept of the logical feature store as a set of capabilities for managing feature data, and the typical design patterns organizations can leverage to support these capabilities.