NengoEdge

High accuracy, low power audio AI. Train and deploy edge devices in the cloud.

Beta users get $100 free credit, and after that you only cover the costs we are charged by our providers for your compute time.

Adaptive control features

Get the best features for your robotics control solution

NengoEdge uses Legendre Memory Units (LMUs), a proprietary recurrent neural network architecture that provides provably optimal representation of time series data. This allows NengoEdge to achieve state-of-the-art accuracy on keyword spotting tasks, as well as other time series applications.

High accuracy

Low power

Low latency

Easy to use

use cases

Use NengoEdge's technology for your voice-enable solution

The first software that allows you to build algorithms from scratch with no code.

Keyword Spotting: Add voice commands to your application by using NengoEdge to detect keywords in a stream of audio data.

Audio Scene Classification: Analyze the ambient audio data around your device to detect events of interest (such as users entering the room).

Anomoly Detection: Detect anomalous audio signals, such as wear and tear in attached machinery.

Use Cases

our story

Why we built NengoEdge

NengoEdge models are designed to be deployed in edge applications with constrained power and memory budgets. NengoEdge automates all the details of efficiently mapping your model on to low power edge devices, taking care of things like weight/activity quantization and op kernel selection in order to optimize power and resource usage.

NengoEdge takes care of all the complicated details involved in training and deploying models for edge hardware. It provides you with an easy to use interface that lets you control the things that matter, while NengoEdge takes care of the device-specific implementation under the hood. This gets you up and running with a minimum of in-house development time. NengoEdge also allows you to rapidly compare accuracy, power, and other metrics across devices, so you can find the best hardware for your application.

Try it now