Introducing MLOps

Chapter 1. Why Now and Challenges

Machine learning operations (MLOps) is quickly becoming a critical component of successful data science project deployment in the enterprise (Figure 1-1). It’s a process that helps organizations and business leaders generate long-term value and reduce risk associated with data science, machine learning, and AI initiatives. Yet it’s a relatively new concept; so why has it seemingly skyrocketed into the data science lexicon overnight? This introductory chapter delves into what MLOps is at a high level, its challenges, why it has become essential to a successful data science strategy in the enterprise, and, critically, why it is coming to the forefront now.

MLOps Versus ModelOps Versus AIOps

MLOps (or ModelOps) is a relatively new discipline, emerging under these names particularly in late 2018 and 2019. The two—MLOps and ModelOps—are, at the time this book is being written, largely being used interchangeably. However, some argue that ModelOps is more general than MLOps, as it’s not only about machine learning models but any kind of model (e.g., rule-based models). For the purpose of this book, we’ll be specifically discussing the machine learning model life cycle and will thus use the term “MLOps.”

AIOps, though sometimes confused with MLOps, is another topic entirely and refers to the process of solving operational challenges through the use of artificial intelligence (i.e., AI for DevOps). An example would be a form of predictive maintenance ...