Practical MLOps

Chapter 12. Machine Learning Engineering and MLOps Case Studies

By Noah Gift

After accompanying Professor Loewi through his procedure I spent more time in his postop care during which he lectured me further. He signed my copy of his little 62-page book, above his signature in a shaky hand he wrote, “Facts without Theory is chaos, Theory without facts is phantasy.

Dr. Joseph Bogen

One of the fundamental problems with technology in the real world is that it is tough to know who to listen to for advice. In particular, a multidisciplinary topic like machine learning is a puzzling challenge. How can you find the right mix of real-world experience, current and relevant skills, and the teaching ability to explain it? This “unicorn” teaching ability is what this chapter aims to do. The goal is to distill these relevant aspects into actionable wisdom for your machine learning projects, as shown in Figure 12-1.

Other domains suffer from the curse of the unbounded complexity that comes with a multidisciplinary field. Examples include Nutritional Science, Climate Science, and Mixed Martial Arts. However, a common thread is the concept of an open system versus a closed system. One toy example of a primarily closed system is an insulated cup. In that example, it is easier to model the behavior of a cold liquid since the environment has minimal effect. But if that same cold liquid goes outside in a regular cup, things get murky quickly. The outside air temperature, humidity, wind, and sun ...