I work at a tech company building libraries for machine learning/statistical analyses, and have done so for the past 10 years since graduating college. I primarily work in R, with the occasional opportunity to work in C++ (though the split is probably 90-10 R to C++). I'd say the role is maybe close to some kind of machine learning engineer, but realistically, I'm not someone deploying models and investigating pipelines, I'm just designing some of the tools to be used in those environments.

I've become increasingly concerned that I might be getting myself stuck in an area of software that isn't particularly marketable. In the MLE space, I think my strength lies in the mathematical foundation for the models in those environments, less in the deployment, and I don't know how important it is realistically to be able to write an SVM from scratch (for example) over just pulling something in from scikit-learn, deploying it, and calling it a day. I have faith that I could probably put together a sensible pipeline on my own as a side project, but that doesn't feel like a replacement for working on something at scale in a real production setting.

Have any of you experienced this kind of tension between the stack you work in and the stack that's actually used in industry? Have you found it to be a barrier to leaving the stack you're in? Is there a point of no return?