Igor Butylenkov

Local LLM inference optimization — MoE, MTP/speculative decoding, quantization, CUDA/llama.cpp