Real Time Voice AI Platform – Gemini Live & LiveKit Integration

Production-grade voice AI platform for hiring and interview workflows, enabling natural, low-latency conversational interactions through Google Vertex AI Gemini Live API with LiveKit-based audio streaming.

  • Built a scalable real-time audio backend using LiveKit integrated with Gemini Live API, enabling natural voice-based conversations with sub-second latency for AI-powered candidate screening and behavioral interview simulations.
  • Implemented LiveKit room and session management to handle concurrent interview sessions with reliable participant lifecycle control and session continuity.
  • Designed bi-directional WebSocket audio streaming to support real-time speech input/output with stable session handling.
  • Deployed compute workloads on Azure Virtual Machines with GCP Cloud Storage for persistent audio/video artifacts and interview data.
  • Achieved 99.5% uptime across concurrent real-time voice sessions, delivering a production-ready Voice AI platform capable of supporting AI-led interviews and interactive hiring workflows.

Tech Stack

Python LiveKit Docker Google Vertex AI (Gemini Live API) WebSockets Azure VM GCP Cloud Storage Multi-Cloud Architecture