VR Interface to Improve Police Training

Research Scientist | Technical Project Lead

I lead a multi-million dollar NSF-funded immersive training platform for emergency responders, integrating LLM-powered non-playable characters, multimodal sensing, and human factors evaluation to build adaptive, measurable, and scalable training systems.

System built
AI-powered immersive VR training platform
Unity-based deployment on Meta Quest 3
Core innovation
LLM-driven adaptive NPC interactions
Behavioral realism with controllable personality traits
Evaluation framework
Multimodal human factors assessment
Behavioral + physiological + usability metrics

The problem

Traditional training lacks adaptive realism
Limited objective metrics of trainee performance
Static scripted scenarios reduce ecological validity

System architecture

Unity-based immersive VR scenarios
LLM-powered NPC dialogue and behavior control
Backend behavioral logging and state tracking

Human factors evaluation

Usability testing under structured protocols
Performance-based behavioral metrics
Physiological sensing for cognitive load assessment

Why it matters

Improves realism and training transfer
Supports measurable performance evaluation
Enables scalable, data-driven training refinement

Technology stack

Unity Meta Quest 3 Large Language Models Behavioral Logging Pipelines Human Factors Experimental Design Multimodal Sensing

Conference Papers