> I can understand this, since most of it was probably written by PhD students; ... many of them without a CS background.
Of the top 4 contributors to scikit-learn, 3 have computer science graduate degrees and 1 has a degree in physics, so I'm not sure that the lack of a "CS background" is the root cause of the majority of the problems with the codebase (perceived or actual).
It may more be related to the nature of academic code in general, since most of it is supposed to be more proof-of-concept rather than general use worthy (e.g., why Google's original code was refactored by Jeff Dean).
Of the top 4 contributors to scikit-learn, 3 have computer science graduate degrees and 1 has a degree in physics, so I'm not sure that the lack of a "CS background" is the root cause of the majority of the problems with the codebase (perceived or actual).
It may more be related to the nature of academic code in general, since most of it is supposed to be more proof-of-concept rather than general use worthy (e.g., why Google's original code was refactored by Jeff Dean).