Scaling models in production is non-trivial. Hopefully this post helps you get some of your research done before setting out on your own scaling journey. TL;DR There's a good reason why everyone says to set your Tensorflow thread counts to 1.

TL; DR: Data scientists should manage their data infrastructure with code. By doing so you can avoid being reliant on engineering departments and also keep better tabs on your costs.

Everyone has their own mental triggers when hearing specific words. Let's take the word "Engineering", which has many triggers, that can range all the way from "Tech shop that writes software" to "Aerospace Manufacturing".

I was an avid AWS user until a year ago, when I started using GCP for various Machine Learning projects.