Chapter 3: From Model to Model Factory

This chapter is all about one of the most important concepts in ML engineering: how do you take the difficult task of training and fine-tuning your models and make it something you can automate, reproduce, and scale for production systems?

We will recap the main ideas behind training different ML models at a theoretical and practical level, before providing motivation for retraining, namely the idea that ML models will not perform well forever. This concept is also known as drift. Following this, we will cover some of the main concepts behind feature engineering, which is a key part of any ML task. Next, we will deep dive into how ML works and how it is, at heart, a series of optimization problems. We will explore how, when setting out to tackle these optimization problems, you can do so with a variety of tools at various levels of abstraction. In particular, we will discuss how you can provide the direct definition of the model you want to...