Boosting Performance: Optimizing Speed in LLM Applications