When software developers talk about artificial intelligence (AI) and machine learning (ML), they mostly mean artificial neural networks (ANN), which are preferably trained and inferred on high-performance enterprise hardware. Today, thanks to tools like TensorFlow Lite (Micro) and Edge Impulse, it is possible to perform at least the inference of a neural network even on embedded systems with limited resources. The charm of this is that the processing of data can take place where the data is actually generated. Examples of this include autonomous driving and automation technology.

Prof. Dr. Michael Stal has been with Siemens Technology since 1991. His research focuses on software architectures for large complex systems (distributed systems, cloud computing, IIoT), embedded systems, and artificial intelligence. He advises business units on software architecture issues and is responsible for the architecture training of senior software architects at Siemens.

In a three-part article series starting August 18, 2023, I will first discuss important AI basics (Part 1), then working with TensorFlow Lite (Part 2), and finally using the MLOps online tool Edge Impulse (Part 3 ). The third part also discusses the AI-supported developer assistant GitHub Copilot, which allows the development of embedded software.