Menu
- (06140) 27, Teheran-ro 27-gil, Gangnam-gu, Seoul, Korea
- +82 (2) 883 1231
The Future of AI
Our vision is to deliver the best AI experience on everything for everyone.
We do this by overcoming hardware constraints with our best-in-class Edge AI Technology
We Make the Impossible Possible with Hardware-aware Machine Learning
As number of edge devices such as smart-sensors, wearable devices,
smartphones and robots grow exponentially, needs for Edge AI are also growing fast.
However, not many can fulfill these needs due to the endpoints’ computational constraints.
We can make this possible.
MCU
We target Microcontroller Unit (MCU) to enable AI on applications such as smart-sensors, wearable devices, robotics, etc.
Our Focus: ARM
CPU
We target Cetnral Processing Unit (CPU) to help clients migrate from expensive GPU instance or to enable Edge AI
Our Focus: Intel
AP
We target Application Processor (AP) to satisfy various needs of smartphone and app makers who need low-power models
Our Focus: Qualcomm
Model-level Compression to
Low-level Optimization
We derive optimum Edge AI model through a full model optimization pipeline,
which is comprised of model-level compression and low-level optimization.
By iterating these two steps, we derive an optimum model output
which could meet our clients’ high standards.
Model-level Compression
Owns custom recipes to compress models up to 0.05% to fulfill clients’ specific tasks
Low-level
Optimization
Accelerated model’s inference time by 2.5x compared to using TFLite for smartphone project
Case Study
Cloud-to-edge migration project with global top medical AI Venture
We supported AIRS Medical to migrate their MRI reconstruction model from cloud’s most expensive GPU instance to local CPU, which eventually solved issued regarding privacy, licensing, marketing and cost
Model Accuracy
Maintained up to 99.7% of our client’s AI model accuracy
Model size
Reduced size of model by 1/1840 (643mb to 350kb)
Inference Time
Accelerated model’s inference time by 160x (9,430s to 59s)
Instance cost
Dropped instance cost to zero by migrating from cloud to edge
Partners
