About Us

Technology

Careers

News

Contact Us

Striving for excellence in
AI NPU software development

Optimizing AI model inference performance and harnessing
remarkable hardware efficiency to pave the way towards world-leading capabilities

BOS NPU AI SW Development Verticals

Providing the most dexterous SW stack and tools
for AI model bring-up and optimization for the world’s leading model inference performance

  • AI Model Compiler
    Development
    (Frontend/Backend)
  • AI Model Development
    & Performance Optimization
  • Runtime &
    Tensor IO API
    Development
  • Quantization
    Algorithms &
    PTQ/QAT Tools
  • AI Model Ops
    Kernel
    Development in
    C++/Python

Full NPU SoC SW Stack

We are developing a full stack SW for top AI model inference performance and utilization

AI Model Development SDK

We are developing all-around AI model development SDK for the easiest and fastest AI model deployment

All-around Programming Model

We are developing NPU SW using two-distinct
NPU SW stacks and optimizing for “Generality” and “Performance”

For “Generality”
Quick deployment and improving performance
  • Out of the box, run many AI models, small overhead incurred
  • Continuous Performance improvement
  • Support for all major frameworks including PyTorch and Tensorflow
For “Extreme Performance”
Complete low-level control of the hardware
  • Targeted AI models running higher performance at higher utilization
  • Completely open hardware allows you to program non-ML applications.
  • Useful for HPC, C++ environments and low-level AI model development.

Source: Official website of Tenstorrent Inc. BOS and Tenstorrent are collaborating on joint AI SW development.

Pre-trained AI Model Compilation Flow

We are developing the most flexible and compatible frontend
& backend AI model compiler for the world’s best AI model inference performance

Source: Official website of Tenstorrent Inc. BOS and Tenstorrent are collaborating on joint AI SW development.