Fundamentals of Deploying a Model as a Service

The purpose of deploying a model as a service is for other people to view and access it with ease, and in other ways besides just looking at your code on GitHub. There are different types of model deployments, depending on why you've created the model in the first place. You could say there are three types—a streaming model (one that constantly learns as it is constantly fed data and then makes predictions), an analytics as a service model (AaaS—one that is open for anyone to interact with) and an on-line model (one which is only accessible by people working within the same company).

The most common way of showcasing your work is through a web application. There are multiple deployment platforms that aid and allow you to deploy your models through them, such as Deep Cognition, MLflow, and others.

Flask is the easiest micro web framework to use to deploy your own model without using an existing platform. It is written in Python...