I think the problem here is two fold: bad UIs and bad APIs. In the science world, there doesn't seem to be any abstraction between backend and frotnend -- the whole thing is one intertwined mess.
Scientists don't always write great software; they write scripts that get the job done quickly in a very specific way. Pair that with the lack of time that they have to wrap things in a GUI and you get a Tkinter/Qt app that is unmaintanable.
I'm currently volunteering with the NZ Department of Conservation. Here is a tiny preview [0] of an app that I have just started making for them. It's purpose is to aid scientists with processing camera trap images using machine learning. I'm building the UI using ElectronJS and Svelte, whilst the backend is a separate Python project.