triton , C++, Ep, Cuda, Python, SGLang, DP, LLM inference frameworks, vLLM, AMD ROCm, ROCdebug-agent, GPU kernel development, hip, rocProfiler, performance optimization, distributed inference strategies, quantization, TP, GPU architecture

triton , C++, Atom, Cuda, Python, KV cache management, Speculative decoding, MXFP4, Transformer and MoE architectures, Quantization, AITER, MoE GEMM, SGLang, vLLM, LLM inference frameworks, FP8, AMD ROCm, GPU kernel development, hip, Sparse attention, Normalization operations