Executing the build
As discussed in Chapter 2, The Machine Learning Development Process, there are several stages we have to go through on the ML project life cycle after performing discovery and building an initial proof-of-concept. These steps are focused on the development of the solution and then the deployment of that solution.
First, we will focus on how we would break down these stages into manageable tasks that could be executed by our engineering team. Each component in Figure 7.2 roughly corresponds to one of these tasks, as follows:
- Prediction Handler / Training Handler: Each of these will consist of application logic that takes a request from the dashboard (via an API request over HTTP) and then triggers the appropriate process. These can be brought together as different endpoints in a simple web service that acts as the interface between the dashboard and the other components of the system.
- Training Pipeline and Forecaster: As discussed in the previous section...
