Welcome to EN:ai

EN:ai Technology

EN:ai enables AI algorithms on different edge devices across IoT boards, FPGA boards and ASIC chips.
Powered by the low complexity and high speedup by tensorization and quantization techniques, we can provide a turnkey solution for edge devices where low-power, high-speed and accurate AI systems are required

Low cost

Powerful AI with different neural network can be deployed on inexpensive IoT/ FPGA boards. And mature edge device design can  further be prototyping to ASIC chip.

Low latency

Edge AI runs locally and reduce the data communication in cloud based system over the air. It will greatly shorten the data accessing and processing time.

Secure data

Data process and inference at local storage and not necessarily send to cloud system. No data will be collected in server.

 

About EN:ai

EN:ai, aims to utilize the emerging tensor arithmetic, added with quantization-aware design flow, for deeply compressing the computational resources needed in the NN hardware design. 

The outcomes from EN:ai will address the fast-blooming needs for AI computing on the resource-limited edge and terminal devices. To enable low-power, high-speed and highly accurate board-level and chip-level implementations, innovative constrained-hardware schemes (including truncated tensor ranks and low-bitwidthquantization) will be devised to facilitate hardware-efficient NN training and inference.

 

The Team

Our team consists of academic and industrial experts specializing in tensor algebra and hardware implementation, as well as business operation.

Max LIU

Executive Director

Ngai WONG

Scientific research advisor

Associate Professor (HKU)

Hao YU

Hardware development advisor

Professor (Sustech)

Edmund LAM

Software & algorithm advisor

Professor (HKU)

Terence LI

VP industrial development
A-CTO (Robotics-Robotics Ltd.)

 

Contact Us

Unit 936, 9/F, Building 19W, No. 19 Science Park West Avenue, Hong Kong Science Park, Pak Shek Kok, N.T., Hong Kong

 
 
Keyboard and Mouse
Projects

Object pose detection

Tiny YoloV3 + CPN

Nvidia 2080ti/ Hi3559/ RK3399pro/ Nvidia Nano

FPGA CNN backbone

EfficiencyNet

Xilinx ultra96 DPU

Quantization/Pruning

Action detection

Pose + RCNN

Fall detection

Defect detection

UNet (segmentation) + ResNet (classifier)

Caffe/Onnx/Pytorch

Nvidia 2080ti/ Hi3559/ RK3399pro

TensorRT/ 8 bits quantization

Image enhancement

Contrast and color boosting

Support up to 7000x5000 pixels resolution

Mobile app

CPU + GPU (general)

Can extend to Video enhancement without flickering problem

Lidar object detection

PointPillars

Vehicles and passengers detection

©2019 by EN:ai Limited