Construction sites are among the most hazardous work environments due to unsafe practices and the improper use of Personal Protective Equipment (PPE). To address these safety challenges, this project proposes an AI-based Construction Site Safety Violation Detection System that automatically identifies unsafe behaviors and PPE violations in real-time video streams. The system utilizes computer vision and deep learning techniques to detect workers, safety gear such as helmets and vests, and hazardous actions including entry into restricted zones and working at heights without protection. A tracking-by-detection approach is employed to monitor individuals across video frames, while pose estimation and action recognition models analyze human posture and movements to classify unsafe activities. When a safety violation persists beyond a predefined duration, the system generates instant alerts to enable timely intervention. This automated approach enhances workplace safety, reduces human supervision effort, and helps construction organizations proactively prevent accidents and injuries.