In contemporary society, the prevalence of posture-related health issues has reached alarming levels, particularly within office and educational environments. In response to these concerns, ergonomics has evolved to include how the individual interacts with the work environment. This study aims to fill gaps in posture correction, using Earables as a pioneering approach to health promotion in a deskwork context. We have developed “EPICS,” a posture inference and correction system combining a multi-sensory earable device and a scalable movement sensor. EPICS can identify seven types of postural movements in real-time, including three poor postures and their transitions through machine learning algorithms. A dual-application mobile platform leverages inference results for immediate inference and feedback. Substantial improvements in posture detection accuracy confirmed the system’s effectiveness, reaching 94% on a 20-subject dataset, showcasing its reliability and robustness to different individuals. Various test scenarios demonstrated the system’s reliability and effectiveness in real-world applications. Survey results indicated that real-time posture correction feedback helps users become more aware of their posture, thus encouraging the development of healthier postural habits, and confirming EPICS’ effectiveness.