Skip to content
SpinorML

Compilers for edge AI · SDKs for vision, LLM and audio

Machine learning on small hardware.

SpinorML builds teenygrad, a compiler that turns Rust into GPU kernels ahead of time. A model leaves your machine as one binary and runs on the device with nothing else installed — no Python, no vendor runtime.

  • Written in Rust, end to end
  • Apache 2.0
  • Running today on NVIDIA Jetson
vision-rs parking garage demoYOLO26 · Jetson Orin Nano Super

The offer

What we do

The compiler

teenygrad

An open source compiler stack that takes a model written in Rust all the way down to GPU code. Free to use under Apache 2.0.

teenygrad.org

Licensing

Commercial licences

The vision SDK is dual licensed. If your product cannot ship under the AGPL, we sell a commercial licence with support.

Licensing

Services

Consulting and training

We take on paid work on GPU kernels, MLIR, cross compilation, and getting models onto embedded targets.

Consulting

The work itself

Everything lives on teenygrad.org

Documentation, benchmarks, the source, and how to get started. Begin there if you want to try it. 

cargo install cargo-teeny
Go to teenygrad.org

Licensing · Partnerships · Support

Contact us

A licence, a partnership, or a hard deployment problem — your message reaches someone who works on the compiler, and you will get a considered answer rather than a ticket number.

* required