PyTorch Helion Hackathon

Join us to push the boundaries of GPU kernel development with fellow developers, researchers, and ML engineers. You'll get hands-on access to Helion, PyTorch's new Python-native kernel authoring language that compiles directly to Triton — making it easier than ever to write high-performance ML kernels.

Helion lets you build roofline-performant kernels with 4x fewer lines of code compared to Triton, dramatically lowering the barrier to advanced GPU programming. In addition, you will get a preview of NVIDIA's CompileIQ, which enables bespoke optimizations and code generation for your kernels leading to higher performance.

Expect deep-dive workshops, mentorship from PyTorch, Meta, and NVIDIA experts, and the chance to prototype real-world kernel optimizations. Whether you're passionate about GPU performance tuning, custom operators, or pushing the limits of what's possible with PyTorch, this is your chance to shape the future of ML infrastructure.

Winning teams will receive prizes including Ray-Ban Meta Glasses, tickets to NVIDIA GTC, and GPUs.