Hey HN! We built an AI recruiting agent that conducts actual voice interviews with candidates over the phone. It sources from 220M+ profiles using semantic sear

Hey HN! We built an AI recruiting agent that conducts actual voice interviews with candidates over the phone. It sources from 220M+ profiles using semantic search (natural language search), then autonomously calls and interviews them. Try it yourself: Call +1 (256) 915-2911 and experience an AI interview firsthand. The agent will ask you screening questions and adapt based on your responses. Takes ~5 minutes. Full platform handles sourcing → calling → interviewing → ranking candidates. Tech stack: - Gemini LLM for conversational intelligence and real-time response generation - Custom speech-to-text + text-to-speech pipeline for natural voice interaction - WebRTC for low-latency audio streaming (open-source components) - ML models for keyword extraction, skill scoring, and behavioral analysis - Real-time orchestration layer that manages interview flow and timing The interesting challenge is making the conversation feel natural at scale. We're handling multi-turn interviews with dynamic follow-ups, interruption handling, and context retention across the entire conversation. Current limitations: - Still struggles with noisy environments - Latency around 800ms could be better - Limitation scoring for soft skills Architecture questions we are curious about: - How would you reduce latency further? We're at ~800ms response time - Better approaches for handling conversational interruptions? - Thoughts on evaluating "soft skills" algorithmically? Call the number and let me know where the AI breaks down or if you'd like to know more.