Machine learning has achieved significant progress in recent years, with systems matching human capabilities in diverse tasks. However, the main hurdle lies not just in training these models, but in utilizing them effectively in real-world applications. This is where AI inference takes center stage, emerging as a primary concern for researchers and