Spinning up serverless infrastructure

Whenever we do any ML or software engineering, we have to run the requisite tasks and computations on computers, often with appropriate networking, security, and other protocols and software already in place, which we have often referred to already as constituting our infrastructure. A big part of our infrastructure is the servers we use to run the actual computations. This might seem a bit strange, so let's start by talking about serverless infrastructure (how can there be such a thing?). This section will explain this concept and show you how to use it to scale out your ML solutions.

Serverless is a bit misleading as a term as it does not mean that no physical servers are running your programs. It does mean, however, that the programs you are running should not be thought of as being statically hosted on one machine, but as ephemeral instances on another layer on top of the underlying hardware.

The benefits of serverless tools for...