How to Build Voice AI Agents That Work in Production | LiveKit × Future AGI

A voice agent isn’t just a chatbot with a speaker attached. In this hands-on session, Jesse from LiveKit and Rishav from Future AGI explore what it takes to build voice AI agents that actually work in production not just in a quiet demo. Jesse builds and deploys a real-time voice agent with LiveKit, covering model selection, turn detection, interruptions, perceived latency, and asynchronous tool calls. Rishav then demonstrates how Future AGI can simulate real-world conversations, evaluate call quality, identify failures, and turn unsuccessful calls into new test scenarios. Topics covered: Choosing the right STT, LLM, and TTS stack Measured latency vs. perceived latency Handling interruptions and detecting when a user is done speaking Avoiding awkward silence during tool calls Testing noisy rooms, accents, bad microphones, and network issues Replaying failed calls and iteratively improving the agent Moving from a working demo to a production-ready voice workflow Resources - Future AGI open-source repo: https://github.com/future-agi/future-agi Try Future AGI: https://app.futureagi.com/ Join the Future AGI Discord:   / discord   #VoiceAI #AIAgents #LiveKit #LLMEvaluation #AIObservability