This project will advance virtual reality–based safety training for construction workers operating near live traffic by developing a scalable, parametric, and AI-enabled training platform that overcomes limitations identified in prior high-fidelity VR deployments. Building on lessons learned from earlier laser-scan–based immersive environments, the proposed effort will create a flexible, rule-based virtual work zone environment using Unreal Engine 5 and deploy it on affordable, portable VR systems to enable broader adoption. The project will integrate data-driven hazard prioritization—leveraging national injury and fatality databases—and incorporate Generative AI to dynamically generate scenarios, simulate worker-to-worker communication, and adapt training conditions in real time. The objective is to deliver a practical, customizable training system that improves hazard recognition and decision-making in live traffic environments, while providing implementation guidance that allows agencies and contractors to efficiently develop new modules aligned with evolving safety needs.