DigiNews

Tech Watch by Johan Denoyer

← Back to articles

Show HN: Agentic CUDA Kernel Optimizer

Quality: 8/10 Relevance: 9/10

Summary

CUDA kernel optimizer using an agentic, LangGraph-based approach to automatically generate, test, and refine GPU kernels. It analyzes workload descriptions, explores launch configurations, and records results to retain the fastest validated implementation. The project is open-source on GitHub with a Python/C++ harness and guidance for setup and execution.

🚀 Service construit par Johan Denoyer