Many applications might benefit from the field of activity recognition, which entails continuous monitoring of the subject’s movements. Human posture recognition is the most critical for this research as it will aid in the creation of highly sophisticated self-guidance systems, such as those created for yoga practice. For example, if the user does not practice the postures with the most exquisite degree of precision, the systems will notify them. The primary objective of this research is to identify several occurrences of yoga poses, extract suitable elements from the human body using Google’s incredible framework MediaPipe Marshall et al., and offer real-time feedback so that the user can fix it. It acts as a prototype for what the author would like to develop as a “personal AI gym trainer.” Besides yoga postures, this system may be used for workouts like dumble lifting, push-ups, and squats. Thus, through MediaPipe I can track the body’s motion in a gym environment, providing a one-stop solution for users following several exercise routines. The method leverages MediaPipe’s excellent pose estimation, which involves extracting crucial anatomical locations on a user’s body. Thereafter, I process this information to detect yoga poses and workouts such as squats, push-ups, and dumble lifting done by the user. The foundation of the system’s ability to provide quick, meaningful feedback on posture correction thanks to accurate real-time body motion identification and tracking.