Planning to create features

A data engineer may build Gold tables from Silver tables for consumption by the business. At the same time, a data scientist is building features from the same Silver tables for models. If we aren’t careful, two people working separately without communication can create different versions of the same metrics. When architecting your unified DI Platform, be sure to think about reusability and maintainability. For this reason, with features specifically, the features-as-code approach is our recommendation. Features-as-code refers to the software development practice everything is code, with a focus on creating a repository of reusable code to define features rather than features stored in tables.

You can implement features-as-code in various ways. Initially, we mainly focus on function reusability. You can place functions you execute in multiple notebooks or scripts in a folder within the repository root directory. In the Applying our learning section...