NVIDIA

NVIDIA's open-source AI infra — Triton Inference Server, NeMo, Megatron-LM, TensorRT. The training and serving stack most production AI runs on.