Evaluating audio quality is a crucial task, with subjective listening tests being the gold standard. However, these tests are time-consuming intrusive and expensive, making them impractical for real-time applications like telecommunications. Despite efforts to develop automatic methods that match human listener fidelity, current approaches have limitations that hinder their use in real-world scenarios. In this paper, we introduce SMARTMOS, a novel approach that addresses some of these gaps and enables accurate, real-time, non-intrusive, privacy aware audio quality assessment. We demonstrate the effectiveness of SMARTMOS through a case study on Noise Suppression (NS) and Packet Loss Concealment (PLC) modules. Furthermore, we show how semi-supervised learning techniques can be leveraged to build a joint MOS model that seamlessly covers both PLC and NS scenarios.