Why Machine Learning Engineers (Or Data Scientists) Are Not The Stars Of The Show

Nothing beats the feeling of implementing a model and training it to the point that it has a high level of accuracy and efficient performance when used on test data (bonus point if your model is small enough to work on edge devices). Not to take anything away from ML practitioners, but customers of commercial product do not utilize your models directly. In fact, they are most likely oblivious to the fact that there's is some form of machine learning involved in the product. And this shouldn't be a surprise. The ML techniques behind Google's suite of tools or Netflix personalization system are not blatantly exposed. There tend to be several interfaces that create an abstraction of the complexity of products.