Voice AI Engineer — Core Cognitics About the role We're building real-time voice agents that power production conversations for enterprises across healthcare, automotive, and scheduling. Our agents handle live calls at scale — which means latency, reliability, and conversational quality aren't nice-to-haves, they're the product. We're looking for an engineer who has shipped production voice AI and understands the hard problems: hallucinations on a live call, multilingual turn-taking, noisy audio, and the tight latency budgets that make all of it interesting. What you'll work on · Building and extending the core voice agent runtime that powers real-time conversations over telephony · Improving conversational quality across multiple languages, including Indic, Arabic, and European languages — real-time language detection, language negotiation, and handling language switches mid-call · Strengthening guardrails across the agent stack: prompt injection defense, hallucination detection, transcription cleanup, turn detection tuning, deterministic tool-calling behavior, and upstream audio denoising · Integrating external systems so agents can take real actions — booking, lookups, updates — across partner APIs in healthcare and scheduling · Working on retrieval and knowledge-grounding pipelines that keep agents accurate and context-aware · Owning production reliability — debugging live customer issues, tracing end-to-end latency across the voice pipeline, and analyzing call quality across staging and production What we're looking for · Production experience building real-time voice or conversational AI systems · Experience with frameworks like LiveKit Agents or Pipecat strongly preferred; comparable experience with Vapi, Retell, or custom realtime stacks also welcome · Strong grasp of the trade-offs between pipelined voice architectures and audio-native realtime models, and when to reach for each · Comfort with LLM observability and metrics-driven debugging (tools like Langfuse, Grafana, Prometheus) · Familiarity with vector databases, embeddings, and retrieval pipelines · Working knowledge of telephony (SIP/VoIP), VAD, noise suppression, and turn-taking · Bonus: multilingual NLP experience (especially Indic or Arabic), MCP familiarity, exposure to regulated domains like healthcare Freshers welcome We're open to recent grads and early-career engineers too — if you've gone deep on voice AI through personal projects, hackathons, open-source contributions, or coursework, we want to hear from you. Show us what you've built: a working voice agent prototype, a retrieval pipeline you tuned end-to-end, latency benchmarks you ran, anything that demonstrates you've actually wrestled with the problems above. Strong project work that shows real engagement with the hard parts of the stack counts as much as job titles. Why join You'll work on voice agents that take real calls for real enterprises — not demos. The problems are genuinely hard, the pace is fast, and you'll own meaningful parts of the system end-to-end.