On Device AI

Ditch the cloud and run AI models directly on device, at the data source with Hardware Accelerated AI. No more latency, bandwidth, or connectivity constraints.


Low Power


On Device AI • Simple

There are many challenges moving AI to the Device. Icii makes it simple. Our Yeti platform creates a Hardware Accelerated AI solution, customized to your model, with only a few simple steps.

AI to Device

Welcome To Yeti

Icii's software, Yeti, automatically transforms your AI model into accelerated AI hardware.

Yeti Demos

Real-time AI On Device

See demos of our initial Yeti build running in real-time in a low power, tiny form factor.

Alpha Testers

Coming Fall 2022

Today, you can sign up for a chance to be selected as an Alpha Tester. Spots are limited.


Icii's Yeti implements AI models on FPGAs. Our focus is high throughput, low power, and small footprint. With Yeti approaching Alpha Testers, we project the following specifications.


High FPS, Low Latency

Continual, fast inferencing is key. We project between 15 to 500+ FPS. One of our pilot ANNs achieves over 300 IPS with less than 3 ms latency.

Low Power

Only 1 to 7 watts

Power is a priority, and the lower the better. With certain FPGAs, we expect to provide AI inferencing under 1 watt.

Small Footprint

6 to 46 sq cm

The future is tiny, and Icii only uses an FPGA and memory. These require only 6 to 46 sq cm of PCB space, sometimes even less.

Icii vs Nvidia

To get a better idea of how Icii performs, take a look at how our projected specs outperform the Nvidia Jetson.