Great points. When first learning data science, it can be easy to get carried away with the most exciting parts, like machine learning algorithms, without focusing on the fundamentals. Ultimately, taking the time to learn best practices will pay off for years down the road. Knowing how to think like a software engineer is an extremely important skill to move into industry as a data scientist.

The top-down approach has served me well in the past. I think a great example of it is the fast.ai series of courses for deep learning. Studying the theory is important, but it can be hard to find the motivation when you’re just starting out. It makes more sense to me to learn how to implement the method and then go back and study the theory in depth.