Nod.AI is a Machine Learning Systems company that provides key enabling technologies such as an open source, MLIR based, high performance drop in compiler and runtime for Machine Learning Frameworks (such as PyTorch via torch-mliror JAX via mHLO). We leverage Machine Learning based methods (Reinforcement Learning in particular) for codegen and auto-scheduling of compute and communication over a heterogeneous set of resources (such as CPUs, GPUs, Accelerators, Systolic Arrays etc). We strongly believe the future is bright for using compilers and to avoid writing hundreds of hand written kernels.
Our A.I Silicon Enablement team has worked on enabling customers from edge devices in the order of milliwatts / microwatts to large dataflow architecture silicon with billions of gates on a chip. If you are building A.I Silicon and don’t have CUDA support talk to us to see how we can accelerate your software efforts. The often underestimated factor in a new A.I Systems company is the effort, cost and time taken to stand up a software stack to rival CUDA – and we are here to help with it.