VR De-escalation Training

I develop Unity and C# systems for a federally funded police de-escalation training platform. The project pairs coordinated multi-user scenarios with a modular, locally hosted speech-to-speech pipeline so sensitive dialogue and performance data can remain inside the lab environment.

Engineering contributions

• Develop locomotion, menus, physical interactions, lighting, and custom VR mechanics for Meta Quest 3
• Prototype shared-scene multiplayer and synchronize character animation with Unity Netcode for GameObjects
• Evaluate Mirror networking for coordinated simulations involving privacy-sensitive research data
• Integrate Whisper speech recognition, an Ollama-hosted Gemma 4 model with approximately 26 billion parameters, and Supertonic voice synthesis
• Structure speech recognition, language generation, synthesis, scenario logic, and performance analysis as independently replaceable modules
• Support contextual prompting, turn orchestration, adaptive NPC state, and structured post-scenario feedback
• Instrument dialogue, emotional progression, tactical outcomes, positioning, physiological measures, system logs, and participant feedback
• Collaborate through Git and GitHub with three undergraduate researchers and two PhD students

System path

The repository is private, so this project is presented as a case study.