Use dbt and FEAST to build a feature store in Teradata Vantage
Overview
This tutorial shows an approach to creating a dbt pipeline that takes raw data and turns it into FEAST features. The pipeline leverages 'ClearScape Analytics functions' for data transformations. The output of the transformations is loaded into FEAST to materialize features that can be used in ML models.
Introduction
dbt
dbt (Data Build Tool) is a data transformation tool that is the cornerstone of the Modern Data Stack. It takes care of the T in ELT (Extract Load Transform). The assumption is that some other process brings raw data into your data warehouse or lake. This data then needs to be transformed.
Feast
Feast (Feature Store) is a flexible data system that utilizes existing technology to manage and provide machine learning features to real-time models. It allows for customization to meet specific needs. It also allows us to make features consistently available for training and serving, avoid data leakage and decouple ML from data infrastructure.