I think part of the problem here is that ML development is extraordinarily more expensive then traditional dev.

I don't generally need to develop my own deployment infrastructure for every new project. However I've yet to see an ml team or company consistently use the same toolchain between 2 projects. The same pattern repeats across data processing, model development, and inference.

Oddly, adding more scientists appears to have a super-linear increase in cost - with the net effect being either duplicated effort or exhaustive search across possible solutions.