Unlock AI potential with ultra-low power technology

We bring you deep learning data processing at sub-milliwatt scale.

Applications

Protected areas

Detect presence or anomalies in natural environments

HVAC Regulation

Optimize HVAC control by detecting occupancy in buildings

Aerial surveillance

Achieve real-time onboard detection in surveillance drones

Atendance measure

Track foot traffic in urban areas

Confined space surveillance

Facilitate presence detection in dark and confined spaces

Product

ASYGN Colibry Neural Processing Unit (NPU) is an ultra-low power microcontroller designed to accelerate AI-driven processing for image, sound, and sensor data. It integrates a fully reconfigurable and efficient Convolutional Neural Network Accelerator (NNPA), enabling high-performance solutions for a wide range of industrial applications.

Ultra-low power Microcontroller

32-bits RISC-V
300 MHz
10 uW – 10mW

Neural Network Accelerator

On-chip AI
Reconfigurable CNN
9.6 GOPS

Optimized for sensor data processing

Video interface
Audio interface
SPI and GPIO

TinyML application ready

640KB SRAM
For data and weight
Always-on AI

Key Features

GOPs

MHz

32-bit

RISC-V

640 KB

SRAM

CNN

Fully reconfigurable

Our chip in your system

Deploy your use cases :

Image classification

From object recognition to anomaly detection, our ultra-low-power chip enables you to run your models at a sub-milliwatt scale for tinyML applications.

Sound classification

Keyword detection and environmental sound classification are now possible at the edge, powered by our ultra-low-power chip that operates at a sub-milliwatt scale.

Motion classification

Ideal for wearables tracking physical activity and security systems detecting motion, our ultra-low-power chip provides continuous performance.

Discover our GREENAI initiative

Features

32-bit RISC-V Core
Up to 300 MHz internal oprating frequency
640KB SRAM
Up to 9.6 GOPS

Reconfigurable CNN Networks
40-pin chip
Between 100 uW and 10mW

ASYGN designs ultra-low power neural processor units to make sensors intelligent and autonomous. From object detection in images to sound recognition, the combination of a RISC-V core and a neural network accelerator delivers efficient processing within just a few milliwatts.