Bringing AI models and inferencing to the IoT

AI computation is shifting from the cloud to the edge to address latency and bandwidth challenges while enabling real-time inference for IoT applications. As computing demands surge at the edge, Synaptics has introduced the Astra SL2600 Series of multimodal Edge AI processors. Designed for high performance and power efficiency, the series supports diverse AI workloads — including voice, vision, and text — to advance the development of the cognitive Internet of Things (IoT).

The processors feature a scalable and secure architecture that adapts to rapidly evolving AI models and algorithms. Through collaboration with Google, Synaptics integrates the open-source Coral NPU and the Torq AI platform, providing flexible computing and an open development environment that helps avoid vendor lock-in caused by proprietary compilers. The architecture combines RISC-V and Arm cores for low-latency, high-efficiency processing with enhanced hardware-level security.

By adopting open-source compiler frameworks such as MLIR and IREE, developers can deploy AI models across platforms more easily, reducing development barriers and accelerating innovation. The long-term Synaptics–Google partnership further advances Edge TPU technology, extending AI capabilities from the cloud to devices. Looking ahead, the integration of CHERI security architecture and quantum-resistant cryptography will make Edge AI even more secure and efficient, ushering in a new era of intelligent, connected devices.

https://www.eenewseurope.com/en/bringing-ai-models-and-inferencing-to-the-iot/