Road accidents remain a global concern, with human er- ror, road infrastructure defects, and environmental fac- tors contributing to millions of fatalities annually. This paper presents RoadGuardian, an integrated multi- modal AI framework designed to enhance road safety through real-time detection of three critical risk factors: driver drowsiness, road potholes, and surrounding ve- hicles. The system employs computer vision techniques with specialized architectures for each detection mod- ule. Drowsiness detection utilizes facial landmark anal- ysis with EAR (Eye Aspect Ratio) and MAR (Mouth Aspect Ratio) metrics. Pothole detection implements a custom YOLO architecture trained on augmented road datasets. Vehicle detection leverages YOLOv8 for ro- bust object recognition. These modules are integrated into a unified dashboard that provides real-time alerts, risk assessment scoring, and situational awareness visu- alization. Experimental results demonstrate high accu- racy rates: 96.8% for drowsiness detection, 94.2% for pothole detection, and 97.5% for vehicle detection with an average inference time of 45ms per frame on stan- dard hardware. The framework represents a significantadvancement in proactive road safety systems, offering a comprehensive solution to mitigate multiple accident risk factors simultaneously.