With the advent of emerging technologies, urban intersections are being increasingly equipped with various types of video-based and in-pavement sensing systems to facilitate round-the-clock monitoring and optimization of multi-modal flows. In comparison, the assessment of the safety performance of these facilities continues to be largely based on either crash history or citizen grievances. Herein lies an opportunity to apply advanced sensing platforms to proactively monitor safety-critical events of multi-modal road users. This work presents a traffic safety monitoring framework which showcases the capabilities of utilizing in-pavement sensors to provide a detailed, automated assessment of mobility and safety-related performance measures for multi-modal traffic at signalized intersections. The term safety-critical refers to any action or interaction that can adversely impact a road user’s safety, including jaywalking, red-light running, and drivers not yielding to pedestrians. To infer the various motorized and non-motorized movements taking place across the intersection, a trajectory-based mode classification algorithm was developed which distinguishes the sensor events on the crosswalk as events triggered by motorized and non-motorized modes. The inferred trajectories are used to analyze safety-critical concerns including driver yielding, crosswalk movement against the red, and driver red-light-running, Conducting this long-term assessment of safety-critical multi-modal traffic dynamics would serve as a framework to harness sensing technologies used for mobility purposes to quantify safety-related performance of intersections.