The convergence between the Internet of Medical Things (IoMT) and physical rehabilitation demands objective monitoring solutions to effectively assess patient progress during therapy sessions. Current approaches often rely heavily on subjective clinical assessments or cloud-based processing architectures, which introduce undesirable latency and significant privacy concerns regarding patient data. This work proposes the development of a low-cost wearable hardware architecture combined with Artificial Intelligence algorithms running directly at the edge, a paradigm known as Edge AI. The system is designed for the acquisition of kinematic data and real-time inference of motor patterns specifically during physiotherapy exercises. Based on a comprehensive systematic review of the literature, the proposed device integrates a modern RP2040 microcontroller with precision BMI160 inertial sensors. This specific architecture aims to overcome the limitations of traditional monitoring systems by offering a non-invasive and energetically efficient solution for evaluating movement quality during sessions, processing data locally to ensure patient privacy and rapid therapeutic response.