Voice AI & Infrastructure Engineer

Role overview Own the conversational quality and production infrastructure of real-time voice agents used in settings where connectivity may be limited. The work spans agent behavior, voice processing, telephony, cloud services, and operational visibility, with the goal of keeping calls responsive and reliable as usage grows. Conversation design includes a focus on natural Hinglish for Indian users.

Responsibilities - Improve conversation flow, turn-taking, and interruption handling across voice agents. - Diagnose and address production issues such as mispronunciations, synthetic speech artifacts, looping, and inconsistent flow execution; add monitoring to catch regressions. - Design prompts and conversation flows for real-time voice APIs, choosing between prompt and architectural changes as appropriate. - Tune the voice pipeline from voice activity detection through speech recognition, language model processing, and speech synthesis to meet a sub-500ms latency target. - Build and operate telephony and inbound or outbound calling capabilities. - Run multi-tenant cloud services, including deployment, scaling, worker readiness, call analytics, latency tracing, and audit trails.

Requirements - Two to four years of relevant experience and at least one real-time voice agent shipped to production. - Strong Python skills and experience owning backend services end to end. - Hands-on experience with a real-time language or voice API, or a speech recognition, language model, and speech synthesis pipeline, including prompt and flow design. - Practical AWS experience with containers, queues, deployment, and operating services. - Ability to investigate non-deterministic failures and determine why an agent looped or produced an unreliable response.

Nice to have - Experience with WebRTC, SIP, telephony, LiveKit, or Pipecat. - Familiarity with MongoDB, event pipelines, or analytics. - Hinglish or multilingual conversation design experience. - Experience with browser automation or scraping, or scaling multi-tenant software.