Data Makes the World Go 'Round

Chapter 16Operationalizing AI

In this chapter, we will examine the following:

• The machine learning life cycle
• What “putting a model into production” means
• Why operationalizing a model is difficult
• What’s involved in putting an AI model into production
• Tools for operationalizing a model

As I noted in Chapter 15, AI is about more than building a model. The value comes when you can operationalize the model by putting it into production as an application or part of a system or workflow. This chapter discusses what is involved in operationalizing AI models and the roles involved. This will give you some insight into the important work that goes on behind the scenes for AI.

The Machine Learning Life Cycle

Putting a model into production is known in some circles as operationalizing the model. Think about it. You may get some insights from a model you build, such as what people are associated with a high probability of a fraudulent transactions or customers who are likely to respond to a campaign. However, increased value comes when you can put that model into production as part of process. For instance, a model to predict fraud should be embedded into systems that process transactions. As the data flows through the model, the transactions can be scored for potential fraud. If it is deemed to have a high probability of fraud, then another action can be taken, such as stopping the transaction.

This is why the machine learning life cycle, which was mentioned in the previous chapter ...