[softwaredoug]

I feel like SWE skills are underappreciated in research code. I've seen a lot of bugs creep in due to poor design or bad testing practices. Leading to the wrong conclusions. Not to mention that its harder for readers to consume if its unreadable code.

Researchers that think their code is "throwaway" dramatically limit their reach.

[cool_man_bob]

It makes sense. I can’t speak for the AI/ML field, but a lot of the software jobs I’ve seen in scientific research areas were pretty obvious they wanted scientists who could do a little code, as opposed to developers who can do a little science.