Technology

Deep analysis of leading edge, multi-modal AI workloads guides every aspect of Recogni's universal AI engine design

Our innovations enable our highly scalable accelerators to achieve the best power efficiency, highest compute density, and lowest total system cost.

Recogni relentlessly innovates so you can deliver your next big thing faster.

Model Development

Train your own AI models or leverage Recogni’s pretrained models
Use Recogni’s SDK to quantize & compile trained AI models onto Recogni’s AI inference accelerator
Utilize Recogni’s bit-exact profiling tools to accurately predict, evaluate & optimize performance

Model Conversion

Achieve high accuracy with push-button PTQ conversion and 100% ONNX operator support
Easily map large multi-modal models across multiple accelerators
Observe the impact of design choices with Recogni’s intuitive profiler

Model Deployment

Recogni's purpose-built universal inference engine (UIE) is the most optimized architecture for accelerating AI inference workloads.

Fully optimized hardware and software architecture, co-designed from the ground up, to optimally accelerate the most commonly used multi-modal AI workloads
Balanced system with the most flexible, power efficient design, with the industry’s highest compute density