Best practices for a new deep learning project

What are the best ways to lay out / structure a machine learning project?

Is this a duplicate issue to this?