NiLabs
Upstream
Where our work lands
Most of what we do goes into other people’s repositories. These are contributions from individual NiLabs members, not partnerships or endorsements.
NVIDIA / OpenShell
A safe runtime for autonomous agents · Rust
vllm-project
LLM inference and post-training · Python
ggml / llama.cpp
Inference on ordinary hardware · C++
Contributing
The most useful thing you can send us is a case where two systems that should agree do not. Open an issue on the relevant repository, or get in touch through our GitHub. No prior experience with inference internals is needed; the good reports usually come from people who hit the problem in an application first.