ML infrastructure tool: the library to share data transformations between Python and Goland parts of the project

Follow the full discussion on Reddit.
In ML world, usually, we have two parts of the project: (part one) training part, where we want to perform a lot of experiments, and we want to have flexible Python-based solutions to fetch data from logs, prepare features, train models... and (part two) applying part; it's our production where we execute our model. Here we are usually focused on performance, high scalability, and stability of code.