> The median data scientist is horrible at coding and engineering in general. The few who are remotely decent at coding are often not good at engineering in the sense that they tend to over-engineer solutions, have a sense of self-grandeur, and want to waste time building their own platform stuff (folks, do not do this).

I can relate to this so much. I've worked in multiple projects with great people who shifted away from solving the problem at hand, to instead construct some sort of generic problem-solving platform. In one project this actually happened twice: after refactoring the beef out of our SpecificProblemSolvingService into GenericProblemSolvingService, the generic problem-solving platform was then rewritten with one extra level of abstraction, so it could run any models designed to solve any task. As far as I know, neither service was ever used to solve any other problem except the SpecificProblem that we were solving the first place.

But it's fun writing platforms, I guess?