AI at the Edge

Chapter 8. Designing Edge AI Applications

The design and development of an application is where all the threads of edge AI are woven together. It requires an understanding of everything we’ve discussed so far, including problem framing, dataset collection, technology choice, and responsible AI. These are in addition to the skills and knowledge required to design a product in the relevant domain and to implement the design in both software and hardware.

In this chapter, we’ll walk through the process of designing an edge AI application, and we’ll learn some of the most important design patterns that are used in real-world applications that can be applied to your own work. By the end of the chapter, you’ll be comfortable with the requirements of edge AI product design and ready to start creating your own products.

There are two main parts to the design of an edge AI application: the product or solution itself, and the technology architecture that makes it work. These two parts are interdependent. The product design will impact the required technology architecture, and the constraints of the technology will influence the design.

In addition, the entire design and implementation process is informed by the specifics of the real-world situation in which the product will be deployed. Your designs will need to evolve fluidly as you collect data, experiment with different approaches, and test your solution under real conditions.

This dynamic system requires an iterative design process, ...