Running models reliably on remote machines

Whether your company uses on--premise infrastructure or cloud infrastructure, you will need to run your machine learning models somewhere other than your laptop at some point. These models might need to be ready to serve fraud predictions, or they may need to process user-uploaded images in real time. You cannot have the model sitting on your laptop and successfully serve this information.

However, as we get our data processing and machine learning applications off of our laptops, we should ensure the following:

1. We should not complicate our applications just for the sake of deployment and scaling. We should keep our applications simple, which will help us maintain them and ensure integrity over time.
2. We should ensure that our applications behave like they did on our local machine, where we developed them.

If we deploy our machine learning...