Source folder

The source folder contains all the core data science/machine learning files. These files are run either on your local machine during development or eventually on a remote compute target.

Util script

Since the core scripts either receive as input a dataset or model and output the same, the util script handles all the credentials to retrieve the artifacts and has functions to retrieve the workspace or the datasets

Variables

All variables (dataset name, features, model name, etc) must be passed as arguments. The advantage of this approach is that you can add a default value in your parameters while developing and, then, once refactoring the project for production, the variables will come from configuration files and sent to the core scripts via arguments.