Dialogue and Event System
Built a modular asynchronous AI dialogue system integrating ChatGPT with JSON-based prompt templates and dynamic interviewer responses.
VR Interaction Framework
Built core interaction systems using Unity's XR Interaction Toolkit, integrating hand tracking, motion controls, and spatial audio for immersive player experiences.
VR Tablet and Quiz System
Designed and developed an in-world VR tablet that appears after the oral interview stage, featuring a fully interactive quiz with UI navigation, input handling, progress tracking, and event-driven state transitions.
Data Recording and Analytics
Implementing pipelines to capture gaze direction, voice activity, response timing, and interaction metrics for AI evaluation.
AI Feedback and Analytics
Developing a generative feedback module where the LLM produces communication and behavioral evaluations after each session.
Scalability and Design Flexibility
Built systems supporting multiple industries and job levels (entry, mid, senior) with customizable scenarios and AI questioning templates.
Profile and LLM Integration
This code configures the AI interviewer based on the player's chosen industry and seniority level (entry, mid, senior). It dynamically builds a system prompt and sends it to the ChatGPT client, which drives the entire conversation. Every player response is sent back to the LLM, and the reply is spoken in-game using Google Text-to-Speech.
AI-Generated Quiz (MCQ)
Instead of hard-coding questions, I use ChatGPT to generate multiple choice questions at runtime based on the selected industry and level.
Sends a strict JSON schema to the LLM to ensure the response is machine-parsable.
Requests a custom number of questions, each with four options and a correct index.
Deserializes and validates the result into a string MCQ list used by the in-game tablet UI.
Voice-Driven Input
This code handles push-to-talk voice input in VR. The system records audio from the microphone, converts it to WAV, sends it to OpenAI's Whisper model for transcription, and forwards the recognized text to the AI interviewer.
Voice-Driven Input
Uses Unity's input system for XR push-to-talk behavior.
Handles microphone recording, trimming, and conversion to WAV.
Integrates with OpenAI Whisper via HTTP multipart form data.
Pipes the transcript straight into the exsting ChatGPT TTS pipeline.