Why do we need model deployment?

If you're already going through the hassle of training and optimizing machine learning models, why don't you take it a step further and deploy it so everyone can use it?

Maybe you want to have the model's predictive capabilities available in a web application. Perhaps you're a mobile app developer who wants to bring machine learning to Android and iOS. The options are endless and different, but all of them share one similarity – the need to be deployed.

Now, machine learning model deployment has nothing to do with machine learning. The aim is to write a simple REST API (preferably in Python, since that's the language used throughout the book) and expose any form of endpoint that calls a predict() function to the world. You want parameters sent to your application in JSON format, and then to use them as inputs to your model. Once the prediction is made, you can simply return it to the user.

Yes, that's all...