This research focuses on improving construction site safety by developing an innovative realtime hazard detection system using machine learning models. At its core is a camera-based backend API that deploys machine learning models for object detection in an SQL database. When potential safety issues arise, the system triggers alerts and logs key data points such as timestamps and alert types. Alerts are grouped into short-term (24-hour) and long-term (1-month) categories, which offer actionable insights for immediate and ongoing safety improvements. By integrating advanced object detection methods like Yolov8 with efficient database management and API-driven functionality, the system offers an effective solution to monitor construction site safety in real time. Its ability to dynamically manage machine learning models and alert categorization functionality make it an essential tool for fostering safer workplaces. Future developments include integrating location data
using compatible sensors to further enhance risk monitoring. Additionally, the paper discusses future integration with the Industry Foundation Classes, which could expand the system’s capabilities by linking hazard detection to specific building elements and processes.