[D] Industry design patterns for fast-moving ML/DL

I have been writing ML code (both training and off-the-shelf model inferences) for close to six years now, but mostly in an academic/personal project setting.

Now, I find myself spearheading an ML project at a big company, and our backend code base keeps growing, and other people depend on it. There are layers to it, with threads spawning, and dependencies on caches and databases for state sharing. It's more than a pet project - you get the gist. I want …

academic backend big big company code code base design industry machinelearning moving patterns people project six training writing