Getting machine learning to production · Vicki Boykis

I haven't seen many posts in the wild on how end-to-end machine learning works, so this post covers the process of creating an end-to-end proof-of-concept (POC) machine learning product, Venti, which is a Medium-like site that generates VC thinkpieces. All the code for serving the model is here, and here for generating inferences. The app, Venti, is live here. A few months ago, Emmanuel Amiesen sent me a copy of his book, Building Machine Learning Applications. Just got this in the mail from @mlpowered and looking forward to digging in (maybe when I'm getting 5 hours of continuous sleep.) It looks great so far and what I like about it is that explains common ML prod jargon like "inference" and "data leakage." I read it with a lot of interest because there is not a lot of "official" literature out in the field about what machine learning orchestration and pipelines should look like.