Selected spaces
A few areas. Pick one.
We focus on a small set of spaces where the work is hard and the feedback is honest. You apply to one; projects form within each space once the cohort is set.
ML Systems
Training infrastructure, kernels, and distributed systems for ML.
Not model-prompting — the engineering underneath. Custom kernels, autograd internals, distributed training, inference serving. Math meets systems.
PythonCUDAPyTorchTriton
Systems & Graphics
Languages and runtimes — and the graphics that run on the GPU.
Build the layers other people write on top of, and the visuals that prove you understand them: parsers, codegen, kernels, shaders, ray marchers, compute pipelines.
RustCWASMRISC-VGLSLWebGPU