I'm looking for resources that talk about best practices or simply some examples (at specific companies or in general) on how tech companies that heavily use machine learning like Twitter, Facebook etc. set up their data science workflow to make it easy to take models from exploratory to production level.
In 2012 Twitter published a paper on what their analytics stack looks like. They heavily rely on Pig. Since this is already a few years old, my quesiton is: Is there a more recent version or something similar to that available for one of the big tech companies using data science?
A recent blog entry from Slack describes their architecture, but it says little about how it is used more specifically for machine learning purposes.