Well, obviously it depends, people can be successful from both camps based on other attributes. But I will say when we interview we often try to weed out people who just took a handful of courses that focus on TensorFlow, but lack general science intuition and depth.

...On the other hand, if you're already a dev and want to become an ML Dev, knowing how to do science deployments, work with big data, and familiarity with APIs like TF would be more valuable than knowing how to do proofs.