From AI Agents to Faster Kernels: Ben Burtenshaw & Felix LeClair (AI Plumbers #2)
Felix LeClair sits down with Ben Burtenshaw from Hugging Face to discuss how the Hugging Face Kernels ecosystem makes high-performance kernels easier to build, share, and run. They cover agent-assisted kernel development, benchmarking results across model families, and why reproducible kernel infrastructure is key for broader hardware support. ## Chapters 0:00 Intro and guest welcome 0:23 What Hugging Face Kernels solves 1:53 Can LLMs help write kernels? 2:06 Using Claude Code + skills to generate kernels 3:21 Benchmarking different models for kernel generation 4:30 Agentic workflows and performance engineering 5:16 Supporting older and affordable hardware 6:55 Why infrastructure matters beyond generated code 7:46 How Kernels Hub helps developers learn and contribute 9:20 How to get involved with Hugging Face Kernels 10:00 Wrap-up Learn more: https://aifoundry.org/ https://huggingface.co https://github.com/huggingface The video was created by @aifoundryorg