The deployment and maintenance of AI applications is more than just a single action; it's a process. In this section, we will work through creating sustainable applications in the cloud by creating a deep learning deployment architecture. These architectures will help us create end-to-end systems: deep learning systems.
In many machine learning/AI applications, the typical project pipeline and workflow might look something like the following:
The training processes are strictly offline, and serialized models are pushed to the cloud and interact with a user through an API. These processes often leave us with several different languages/packages/tools talking to each other in a way that slows down our processes. Often, the model creation part of this process becomes the domain of data scientists, while the deployment and maintenance part becomes that of an engineering...
