Neuton is a no-code TinyML platform that automatically generates compact machine learning models, typically under 5 KB, and embeds them into microcontrollers and sensors without compromising accuracy. This technology enables the deployment of AI-driven functionalities in low-power edge devices, addressing the challenge of integrating advanced analytics in resource-constrained environments.
Funding
Funding not disclosed
Founders
Product
Problem
Integrating AI-driven functionalities into resource-constrained edge devices like microcontrollers and sensors is challenging due to the limited processing power and memory. Existing machine learning models are often too large and computationally intensive for these environments.
Solution
Neuton offers a no-code TinyML platform that addresses these challenges by automatically generating extremely compact machine learning models, typically under 5 KB, designed for direct deployment on microcontrollers. The platform's patented neural network framework forgoes error backpropagation and stochastic gradient descent, growing the network structure neuron by neuron. This approach enables the creation of models with minimal size and high accuracy without requiring additional compression techniques. The platform supports 8, 16, and 32-bit microcontrollers, allowing developers to embed AI-driven capabilities into a wide range of edge devices.
Target Audience
The primary audience includes IoT developers, embedded systems engineers, and AI researchers seeking to deploy machine learning models on microcontrollers and other edge devices with limited resources.
Features
- No-code automated TinyML platform with a patented Neural Network Framework
- Automated pipeline builds compact and accurate models without manual tuning
- Models average less than 5KB, enabling deployment on resource-constrained devices
- Supports 8, 16, and 32-bit microcontrollers
- Unique algorithm forgoes error backpropagation and stochastic gradient descent
- Automatic neuron-by-neuron structure growth optimizes model size and accuracy
- Constant cross-validation enhances the generalizing capabilities of the model
- Seamless integration into ultra-low power smart sensors