The exponential growth of the fitness sector has generated a critical gap between mass access to strength training and the availability of specialized technical supervision. A significant proportion of practitioners exhibit deficiencies in fundamental movement patterns, contributing to a high number of sports injuries and reducing training effectiveness. This work presents REPS (Rating Exercise Performance System), an automated system for exercise technique analysis and correction based on computer vision and artificial intelligence. The system integrates artificial intelligence technologies such as MediaPipe for three-dimensional pose estimation and DeepSeek for personalized feedback generation, implementing a processing pipeline that includes algorithms for: automatic repetition detection through temporal signal analysis, intelligent synchronization with expert patterns using adaptive interpolation, anthropometric normalization through affine transformations to compensate for body differences, and skeleton alignment for frame-by-frame comparison. The hybrid evaluation methodology combines universal metrics applicable to all exercises (amplitude, symmetry, trajectory, velocity) with specific metrics adapted to each particular exercise. The system processes training videos to generate detailed biomechanical analyses through objective scores, comparative visualizations, and contextualized textual feedback. Results demonstrate that the system can successfully process videos of fundamental strength training exercises, generating comprehensive performance assessments and personalized recommendations.