sybil-solutions/inference-optimisation-skill: An agent skill that tunes local LLM inference for your hardware

github.com5 min read

An open-source agent skill uses hill-climbing optimization to tune local LLM inference performance across diverse hardware and quantization settings. The tool calculates theoretical performance ceilings, iteratively measures levers against interleaved controls, and enforces stability and memory constraints before publishing qualified builds. Open article →