Google Research and Synaptics have launched the Synaptics Coralboard, a development platform designed to accelerate the creation of edge AI products.

The Coralboard is powered by the Synaptics Astra SL2610 product line and features the Synaptics Torq neural processing unit (NPU) alongside Google’s Coral NPU technology. It is intended for power-efficient, always-on applications that require multimodal on-device AI capabilities.

The hardware specifications include a 2GHz dual-core system-on-chip (SoC) with 2GB of DDR4 DRAM and a 1 TOPS NPU subsystem capable of handling convolutional neural networks and transformer models. Connectivity options include CSI camera input, DSI display support, and interfaces for microphones and sensors. Developers can expand the platform via M.2 modules for Wi-Fi and Bluetooth.

At the recent Google I/O event, the companies demonstrated “Jellectronica,” an AI-powered music experience. The system uses an object detection model to track movement from a live aquarium stream, converting the data into control signals for a generative music performance powered by Google DeepMind’s Lyria model.

“The Coralboard is designed to give developers and OEMs a fast path to build and deploy real-time, multimodal AI products for the edge at scale,” said Vikram Gupta, General Manager of the Edge Compute division at Synaptics.