An AI Engineer Technical Guide to Feature Store with FEAST

Dive into a comprehensive technical guide on Feature Store implementation using FEAST (Feature Store). Learn about the key data challenges in productionizing ML systems and how feature stores address them. Explore FEAST's capabilities as an open-source feature store, enabling on-demand transformations and combining request data with precomputed features. Follow along with a hands-on Jupyter Notebook demonstration covering FEAST installation, source data understanding, registry catalog and online store setup, and architecture review. Discover how to transform feature values, work with online and offline stores, and validate features. Gain insights into machine learning with online features, model saving, and scoring using historical data. Conclude with a content review, GitHub repository exploration, and future plans for using PostgreSQL as an online store.

Video Start
Feature Store content intro
Feature Store - What is it, and how it helps?
Feature store - Details
Google Feature Store - Vertex
DataBricks Feature Store
Tecton Feature Store - FEAST
Feature Store Definition
Jupyter Notebook: Feast Installation/Init
Understanding Source Data
Setting Feature Store - Creating registry catalog and online store
Feast Architecture Review after hands-on example
Online store sqlite review
Transforming the feature values from source data
Understanding Online and offline store
Features added to online store validation
Machine Learning with online features
Saving Model
Using historical data and saved model to score
Content Review
GitHub review to Jupyter Notebook
Plans to use Postgresql in place of sqllite as online store
Credits