Reliable Machine Learning

Chapter 12. How Product and ML Interact

As companies rush to use ML capabilities to meet customer needs, they are eager to leverage cutting-edge research to tackle a wide variety of business applications. Many of the product teams and business managers, still anchored in traditional software product development methodologies, find themselves in a new and unfamiliar territory: building ML products.

Building your first ML product can be overwhelming. It’s not just a question of getting ML right, difficult enough in itself; rather, the integration of the ML into the rest of the product (and the rest of the business) requires many things that need to work together. Among these, data collection practices and governance, the quality of data, definition of product behavior, UI/UX, and business goals all contribute to the success of an ML-based product or feature.

Different Types of Products

One of the important and useful features of ML is that it can be applied to many types of products. It can be used in analytics applications to derive insights about business trends and metrics. It can be rolled into an appliance or device to be shipped to consumers. Sophisticated ML systems are built into self-driving cars to detect other objects and to make decisions about driving. The breadth of ML applications is huge and growing. As a result of this huge diversity of use cases, organizations focused on integrating ML into their existing or new products face an extremely steep learning curve ...