Project Overview: We are building a real-time AI Female Tutor companion using React Native. We are using the xAI Grok Voice Agent API (via LiveKit) for the voice and brain. We need an expert to build a fully client-side 3D avatar that performs real-time lip-syncing (visemes) based on the audio stream. CRITICAL REQUIREMENT: We will NOT use cloud-video streaming services like HeyGen, D-ID, or Tavus due to high per-minute costs. The avatar must be rendered on-device (client-side) using the phone's GPU. Scope of Work: Grok/LiveKit Integration: Implement the Grok Voice Agent API using the LiveKit React Native SDK. Client-Side 3D Rendering: Integrate a realistic 3D female tutor (Ready Player Me or GLB format) using React Three Fiber (R3F), ViroReact, or Unity-as-a-Library. Real-Time Lip-Sync (Visemes): Create a system to analyze the raw audio buffer from the Grok stream and map it to the 3D model’s Morph Targets/Blend Shapes (MouthOpen, Pucker, etc.) with sub-100ms latency. Character Behavior: Implement idle animations (blinking, breathing, slight posture shifts) to ensure the tutor looks "alive" even when silent. Performance Optimization: The solution must maintain a high frame rate on standard iOS and Android devices without overheating the hardware. Required Technical Skills: Expert React Native Developer with experience in Native Modules. 3D Graphics Specialist: Deep knowledge of Three.js / React Three Fiber or Unity C#. Audio Engineering: Experience with WebRTC audio tracks and FFT/RMS analysis for lip-syncing. Asset Management: Experience with Ready Player Me avatars and ARKit-compatible blend shapes. Deliverables: A clean, modular React Native component for the 3D Avatar. The bridge logic linking Grok's live audio to the avatar’s facial movements. A fully functional demo: The user speaks, Grok responds, and the 3D tutor talks back with perfect lip-sync.