This chapter covers:
- Embedding your models into the system you are going to build
- Dealing with nonfunctional implications
- Building the data and model-serving infrastructures for production
- Ensuring that the user interface is appropriate
- Ensuring that the logging, monitoring, and alerting elements are properly governed and managed in production
In sprint 2, the team built, tested, and selected models to support the user stories developed in sprint 1. Without more work, the models cannot be used to generate value; essentially, they’re just lines of code sitting inanimate in a repository. To be useful AI, models need to be implemented in the IT architecture that supports the client’s business processes and customer delivery.
There are two characteristics of an ML system, which are a focus of the work in sprint 3, the topic of this chapter. We must implement the data infrastructure supporting the production platform to ensure that the models operate in the data environment that they were trained for. Additionally, we need to retrieve and run the ML models created from sprint 2 for the system to work.