Bolt on the toolkit

The fleet's standard AI toolkit is PyTorch. Bolt your Mojo conv op onto it so any crew member can call it from a torch model like a built-in, no Mojo knowledge required.

A PyTorch custom op registers your Mojo/MAX kernel under a torch namespace. Once bound, it flows through torch tensors and composes with the rest of a model, how production teams ship custom CUDA-class kernels.