In this chapter, I described a high-level architecture and approach to design a data-driven enterprise. I also introduced you to influence diagrams, a tool for understanding how the decisions are made in traditional and data-driven enterprises. I stopped on a few key models, such as Kelly Criterion and multi-armed bandit, essential to demonstrate the issues from the mathematical point of view. I built on top of this to introduce some Markov decision process approaches where we deal with decision policies based on the results of the previous decisions and observations. I delved into more practical aspects of building a data pipeline for decision-making, describing major components and frameworks that can be used to built them. I also discussed the issues of communicating the data and modeling results between different stages and nodes, presenting the results to the user, feedback loop, and monitoring.

In the next chapter, I will describe MLlib, a library for machine learning over distributed set of nodes written in Scala.