I am a postdoctoral researcher at the University of Central Florida’s Smart & Safe Transportation Lab. My research focuses on AI-driven transportation safety, connected-vehicle trajectory and video analytics, proactive crash prediction, causal and behavior-aware safety modeling, and safe integration of ADAS/CAV technologies into mixed traffic.
My work aims to move transportation safety research from after-the-fact crash explanation toward real-time, interpretable, and actionable risk evidence for agencies, infrastructure operators, and emerging mobility systems.
B.Eng., M.Eng., and Ph.D. in Traffic and Transportation Engineering
School of Transportation, Southeast University
I develop safety analytics that connect high-resolution traffic behavior with interpretable crash-risk evidence. My work spans connected-vehicle trajectories, traffic video, causal inference, Bayesian and random-parameter models, LLM/VLM-enabled event understanding, and ADAS/CAV safety evaluation.
Transforms CV trajectories, risky-driving events, and traffic states into real-time crash risk and crash type predictions.
Representative work: TR-C 2025 causal mediation framework with connected vehicle data; under-review BiLSTM-Transformer freeway crash risk model.
Moves from prediction accuracy to mechanisms, counterfactuals, and policy-relevant explanation for roadway safety decisions.
Representative work: causal mediation, double machine learning, causal forests, and interpretable ML for crash risk, ADAS, and warning systems.
Models how crash risk and injury severity vary across time, space, intersections, counties, road users, and traffic contexts.
Representative work: AMAR 2025 grouped random-parameters Poisson-Lindley model with spatial effects.
Studies pedestrian, motorcycle, moped, and right-turn interaction risks using crash records, conflict data, field studies, and behavioral evidence.
Representative work: AMAR 2025 pedestrian injury severity across vehicle movements; AAP 2024 distracted pedestrian crossing safety.
Evaluates the real-world safety promise of ADAS/CAV systems, driver warnings, platoons, and cyberattack impacts in mixed traffic.
Representative work: ADAS effectiveness, in-vehicle warning personalization, and connected platoon cyberattack risk assessment.
Builds toward first-person near-miss/crash datasets and LLM/VLM/diffusion workflows for interpretable traffic video safety analysis.
Representative work: SAVeD dataset under review; LLM-diffusion framework for ADAS crash event understanding in preparation.
I am especially interested in collaborations on connected-vehicle analytics, crash-risk prediction, causal and Bayesian safety modeling, ADAS/CAV evaluation, vulnerable-road-user safety, and multimodal video understanding for near-miss and crash events.