Edge AI
We used Neural Networks for decoding encrypted Wideband radio communication 30 years ago, since then both hardware and network topology has evolved.
Based on our experience with amongst other the Nvidia solution stack, we decided to investigate the possibility to develop simpler but still capable, more energy efficient and very robust offline AI concepts (Edge AI).
By combining our experience in electronics, radio communication, FPGA and neural networks, we are prototyping AI systems for smart sensors.
The article “Attention is all you need” from 2017 and the following success of Transformer based AI Large Language Models (LLDs) as well as the interesting Global Workspace Theory (GWT) also makes small scale AI implemented in RISC-V technology very interesting.
Some of the tools we use:
