Voice AI Infrastructure & Integration Engineer

You will build the monitoring and tracing stack that gives full visibility into every call, own the call recording pipeline that satisfies FDCPA compliance, scale the orchestration runtime across container infrastructure, and integrate with telephony providers for SIP trunking, call routing, and audio stream handoff.\n\nKey Responsibilities:\nObservability, Logging & Monitoring\nDesign and implement the full-stack observability pipeline using OpenTelemetry, Datadog, and ELK Stack — distributed traces, structured logs, and real-time metrics across every layer.\nInstrument the voice pipeline to emit per-turn telemetry: STT latency, LLM inference time, guardrail check duration, TTS first-chunk latency, and total round-trip time.\nBuild alerting rules for latency SLA breaches, guardrail trigger rate spikes, STT/TTS error rates, and telephony drop rates.\nCreate operational dashboards for real-time call volume, concurrent sessions, and per-vendor health.\nCall Recording & Compliance\nOwn the call recording pipeline: capture full-duplex audio and per-turn transcripts, store in S3 / Azure Blob with FDCPA-compliant retention.\nIntegrate with the QA & audit portal (MaestroQA / EvaluAgent) for call playback, AI-assisted scoring, and compliance flag review.\nEnsure PCI-DSS compliance for stored audio: redact payment card segments before long-term storage.\nCI/CD, Deployment & Scaling\nBuild CI/CD pipelines for the voice agent codebase — orchestrator configs, system prompts, Markdown scripts, guardrail policies, and infrastructure-as-code.\nContainerize voice pipeline components using Docker and deploy on Kubernetes (EKS / AKS / GKE) with auto-scaling based on concurrent call volume.\nImplement blue-green and canary deployment strategies for zero-downtime rollouts.\nManage session memory infrastructure (Redis / Upstash) with TTL, expiry, and failover for per-call state.\nTelephony & Post-Call Integration\nIntegrate with Twilio, Vonage, and Plivo for SIP trunking, call routing, and PCM / μ-law audio stream handoff.\nConfigure calling windows, timezone management, contact frequency caps, and DNC enforcement at the telephony layer.\nBuild post-call async jobs: CRM update, email confirmation, call summary generation, and payment gateway integration.\nImplement human escalation handoff — warm transfer with full conversation context passed to live agents.\n\nQualifications:\n6–10 years in DevOps, SRE, or platform engineering, with 3+ years supporting real-time or voice/telephony systems.\nBachelor’s or master’s in computer science, Information Systems, or related field.\nTrack record building production observability stacks for distributed, latency-sensitive systems.\nDeep Kubernetes expertise — auto-scaling, GPU node management, multi-service deployments.\nExperience with telephony protocols (SIP, RTP, WebRTC) and at least one CPaaS provider.\n\nPreferred:\nExperience in BPO, contact center, or financial services voice AI infrastructure.\nHands-on with GPU infrastructure management (NVIDIA A100/H100, Triton Inference Server).\nFamiliarity with voice pipeline orchestrators (Pipecat, LiveKit, Vocode) and their scaling patterns.\nExperience with per-call cost optimization for real-time AI systems.