We are looking for an experienced Node.js Engineer to build a critical business feature: a real-time AI Voice Agent capable of autonomous outbound calling. This is a technical product ownership role where you will architect the system from the ground up using Node.js.You will be responsible for solving the "hard problems" of voice AI: latency optimization, handling user interruptions ("barge-in"), and noise cancellation. You will lead the implementation of our first non-English language (Dutch), working alongside native speakers for quality validation, and design the architecture to support future languages.
Main
Responsibilities
Voice Agent Development (Node.js): Build and maintain the core voice agent logic using Node.js, ensuring sub-second latency for a natural conversational feel.AI & Telephony Orchestration: Integrate Speech-to-Text (STT), LLMs, and Text-to-Speech (TTS) services (e.g., Deepgram, OpenAI, ElevenLabs) with telephony providers (e.g., Twilio, Vonage) into a seamless event-driven architecture.Latency & Interruption Handling: Implement sophisticated logic to handle "barge-ins" (when a user interrupts the bot) and silence detection, ensuring the bot stops speaking immediately and listens, just like a human would.Multi-Language Architecture: Design the codebase to support internationalization (i18n), starting with Dutch.
Note: You do not need to speak Dutch; you will work with a Product Owner who handles linguistic testing.
Infrastructure Selection: Evaluate and recommend telephony and Voice API providers if the current infrastructure is insufficient for production-grade AI voice calls.
Requirements
Expert in Node.js: Strong backend experience with Node.js/TypeScript, particularly with WebSockets and Streams (essential for handling real-time audio data chunks).
Voice AI Experience: Proven experience building voice bots or conversational agents. You understand concepts like VAD (Voice Activity Detection), STT/TTS latency, and SIP trunking.
API Integration: Deep experience orchestrating multiple async APIs (LLM + Voice) simultaneously without blocking the event loop.Problem-Solving: A "scars over certificates" mindset. You can describe specific instances where you fixed audio quality issues, reduced latency, or handled edge cases in phone conversations.We offer:Remote work modelFlexible working hours45 working days per year for personal time outMentorship and knowledge-sharing
Originally posted on Himalayas
