Agentic CUDA Kernel Optimizer

Agentic CUDA Kernel Optimizer is a developer tool that automates the optimization of CUDA kernels by iterating through code generation, benchmarking, correctness checks, and refinement. It uses a LangGraph workflow and a standalone C++ harness to compile and launch kernels on NVIDIA GPUs, leveraging OpenAI models for guidance and Nsight Compute profiling for performance insights.