Soo, so true.

Working with Tensorflow, python is unfortunately the only realistic option. I'm getting the hang of it, but I'd still love to use ruby for it – it's all just data wrangling, it doesn't even need to be fast, and ruby code is just beautiful (to my eyes, I know it's a matter of opinion).

I've also only recently discovered the beauty that is rake. Especially for data pipelines it's a fantastic tool. Hit a bug? No need to rerun everything – it'll pick up at the exact step that failed. One of ten data files changed? It knows exactly what needs to be redone. Etc.