Vlad ButacuSolutions architect · AI systems consulting

Blog

Night desk with a monitor showing llama.cpp flags, a notebook on memory budget, and books on llama.cpp and attention.

Latest

Your Local LLM Is Slow Because of Five Config Flags

KV cache, context size, and batch size often matter more than the GPU. llama.cpp and Ollama parameters on consumer hardware.

6 min read

Tell me what you are building.

The call is free and carries no obligation.

Send an enquiryBook a 30-minute call