E-SMART powers Speed Management for Truck Fleets with
DeepliteRT Speed Sign Detection on an Arm Cortex-A53 CPU
Execution time of 109 ms; 3.77x faster than ONNX-RT
Person Detection on NUCLEO-H743ZI Arm Microcontroller @280 MHz
Yolov5n (7 MB) trained on COCO-person dataset
Deeplite Neutrino compression plus INT8 reduced memory requirement to 396 KB

AI-Driven Optimization to make Deep Neural Networks faster, smaller and energy-efficient from cloud to edge computing

Make AI accessible and affordable to benefit everyone's daily life.

Enable Edge Computing

Create new possibilities by bringing AI computation to every day devices such as cars, drones and cameras.

Economize on Data Centers

Faster DNNs lowers the costs on cloud and hardware back-ends, helping businesses scale their AI services.

Faster Time-to-Market

Automated design space exploration can drastically decrease development efforts by easily finding robust designs.


We combined breakthrough AI technology with years of research at the world's top universities to create Deeplite Neutrino™, a powerful and scalable hardware-agnostic software accelerator.

Introducing Neutrino™

Deeplite Neutrino™ leverages a novel multi-objective design space exploration approach to automatically optimize high-performance DNN models, making them dramatically faster, smaller and power-efficient without sacrificing performance for real-time and resource-limited AI environments.

More power efficient

Neutrino™ provides power API to seamlessly fit into your routine workflow with minimal effort. The engine is designed to be intuitive and integrated with existing AI frameworks.

Design Space Exploration

Neutrino™ delivers a fully automated, multi-objective design space exploration with respect to operational constraints, producing highly-compact deep neural networks.

Smart Quantizer

Neutrino™ exploits low precision weights using highly-efficient algorithms that learn an optimal precision configuration across the neural network to get the best out of the target platform.


Accelerate computer vision ML inference with high-performance, 2-bit quantization runtime for Arm Cortex-A CPUs.

· Deploy advanced video analytics and computer vision features on cost-effective Arm CPUs.

· Faster time-to-market and compatibility for existing mobile devices, surveillance cameras, and machine vision systems.

· Lower-cost hardware solutions than developing custom GPU or NPU hardware designs.

Smart 2-Bit Quantization

World leading model optimization that leverages training aware 2-bit quantization to retain model accuracy while reducing memory bandwith.

Arm CPU Runtime

Fastest Arm Cortex-A Neon runtime. Provides compute optimization that delivers the highest inference performance and power efficiency.

PyTorch Optimization

Optimal deployment path for PyTorch vision models for Arm-based embedded systems.


Our result is a robust AI on cost-effective hardware platforms. Our customers use our solutions to maximize their investments in AI experts and scale their deep learning development with one standard software.


Real-time deep learning on low-power processors to help move people and things around the planet safely.

Learn More

On-device AI for an engaging experience that captivates the user and conserves battery life.

Learn More
Smart Cameras

Understand, analyze and make predictions from video and images in an instant using real-time deep learning.

Learn More
IoT Devices

Deploy deep learning on edge devices and smart sensors to generate new insights and intelligence.

Learn More