Institute for Communication Technologies and Embedded Systems

EXIST Research Transfer: Development Tools for Neuromorphic AI Systems


Neuromorphic computing systems offer disruptive performance improvements in the field of artificial intelligence (AI). Neuromorphic chips already available today promise 10-100x improved energy efficiency, latency and space efficiency compared to conventional technology. They are particularly strong in inference "on the edge", i.e., the application of trained AI models that are executed directly on the device. The increases in performance are achieved through a paradigm shift in chip hardware. While AI applications are currently executed on traditional CMOS hardware with separate computing and memory units, neuromorphic computing systems emulate the human brain, in which the computing and memory units are combined.

However, the available technology is not in widespread use today, as neither semiconductor manufacturers as chip suppliers nor AI device OEMs as users have the necessary hardware and software development tools for system integration. The reason is that both hardware integration (of the chips) and software design differ significantly from existing methods.

This project bridges the gap between novel, neuromorphic chips, and their use in disruptive AI products. Our Software Development Kit (SDK) enables the programming of AI applications on neuromorphic systems and simulates applications before neuromorphic chips are ready for series production. Our Hardware Development Kit (HDK) integrates purchased neuromorphic chips into a test system and enables the validation of neuromorphic hardware. We are developing the fully integrated product, with the SDK forming the core. We are harnessing the potential of neuromorphic technology at an early stage, thereby making an important contribution to sustainable economic development in Germany and Europe.