Practicing Trustworthy Machine Learning

Chapter 8. An Ecosystem of Trust

So far in this book, you have learned about the tools required to embark on the journey of becoming an effective responsible ML practitioner. But are they enough? What else do you need to build, deploy, and iterate on ML systems that a diverse group of stakeholders can trust? Beyond the individual tools and techniques covered earlier, you need to know how to put these different pieces together to reach an ML-driven solution for your business problem while minimizing downstream risks. This sounds daunting at first, but there are resources in this growing field that can help achieve this goal.

This chapter will cover tools, tactics, and frameworks that provide a bird’s-eye view of what’s going on within and across ML models inside a company. To begin with, you’ll learn about technical tools for implementing ML pipelines and about guidelines and strategies to navigate the human-in-the-loop steps in ML pipelines effectively. You’ll be introduced to a few concepts and resources that help gain a cross-project outlook on ML workflows inside your company. Finally, you’ll see how all this knowledge can come together by exploring a deep-dive example of implementing an ML-based recommender system. If you are a product or engineering leader, the resources in this chapter will help you effectively collaborate with business stakeholders on implementing trusted ML. If you are an ML engineer or data scientist, you will gain valuable context for understanding trade-offs ...