Real Time AI Contact Center Platform

I am ready to move a production-grade, real-time contact-center platform that runs on Asterisk v18/21. The goal is to handle both inbound and outbound calls while keeping conversational latency under 800 ms at the 95th percentile, even when 300+ sessions are live.

The heart of the system is an AI layer that must combine speech recognition, natural language understanding, and speech synthesis so callers can interrupt (barge-in) naturally and still maintain contextual continuity. Low-latency, full-duplex audio streaming is essential, and the code you deliver should expose clear hooks for campaign management, operator escalation, and horizontal scaling.

Operational visibility is just as critical: metrics, structured logs, tracing, and an admin console all need to be in place so I can meet or exceed 99.5 % uptime. Everything should fit cleanly into our existing CI/CD pipeline, and your solution must remain cloud-agnostic so we can scale on-prem or in AWS without heavy refactoring.

Deliverables
• Asterisk-based contact-center micro-service stack (Docker/K8s preferred)
• Real-time AI dialog orchestration engine with barge-in support
• API/SDK for campaign creation, operator takeover, and reporting
• End-to-end observability (Prometheus/Grafana dashboards, log/trace exporters)
• Load-test scripts proving ≤800 ms p95 latency at 300+ concurrent calls
• Deployment guide and hand-off documentation

Acceptance criteria
1. A 4-hour soak test shows ≤0.5 % error rate and no latency breaches.
2. Voice interactions demonstrate smooth barge-in and context retention.
3. All metrics and logs are viewable in the provided dashboards without manual tweaks.

If you have deep Asterisk knowledge, real-time audio experience, and a passion for AI-powered voice applications, I want to see how you would bring this platform to life.