Have you wondered what goes into building a REAL WORLD tinyML device/product? In this series, we discuss the details of how product developers and engineers built their tinyML devices, from early development phases to commercialization. The discussion will be a deep dive into engineering and technologies that these teams have developed/utilized and lessons learned.

During this talk we will discuss the practicalities of building a complete solution, ready to manufacture, incorporating TinyML. We will center the discussion around the development of two solutions from Reality AI, a Renesas Electronics company. One of them is Automotive SWS (where SWS stands for “See with Sound”), a solution for the automotive and mobile robotics sectors. SWS uses sound to supplement camera, lidar or radar systems for collision avoidance in vehicles, and is available to incorporate into automotive systems. The other is RealityCheck HVAC, a hardware, software, and TinyML solution for creating self-diagnosing air conditioners and heat pumps.

Topics will include how ML fits into the product development and engineering lifecycle, why constructing the machine learning model is only a small part of the total cost of completing a project, and the solutions available from Reality AI / Renesas to use the output of the TinyML model building process to optimize the Bill of Materials and hardware design. These subjects will be illustrated using Reality AI’s multi-year efforts to create Automotive SWS and RealityCheck HVAC as examples.