I reduced LLM inference GPU calls by 94% using semantic routing — Score 11/100 · Other | indexator.ai | Indexator.ai