Voice AI Implementation Contractor for Contact Center Automation (Five9 / Retell / Vapi / Similar)

We are launching a new conversational voice-AI capability in an established contact-center environment. The immediate work is to deliver customer self-service flows such as appointment scheduling and payments, integrated with our existing telephony and business systems. The target call architecture is explicit: Five9 remains the call anchor. Calls arrive over existing private carrier/SIP connectivity, are handed to the voice-AI platform through SIP REFER, and return to the appropriate Five9 skill/queue when a live agent is needed. The contractor must be able to design, implement, test, and troubleshoot that end-to-end handoff, not only build prompts or dialog flows. Our chosen platform is Kore.ai, but we are NOT looking for a Kore.ai specialist. We want a senior, hands-on builder who has successfully delivered comparable production voice agents on platforms such as Retell, Vapi, Bland, PolyAI, Cognigy, Google CCAI, Amazon Connect, or similar. You will be the primary implementation lead for the voice-AI layer, partnering with an internal team that already owns Five9, applications/integrations, and operations. This is a long-term, project-based contractor role with the potential for ongoing work as new use cases are added. What you’ll do - Turn business requirements into production-ready voice-agent designs and call flows. - Design conversation logic for scheduling, payments, verification, common questions, corrections, no-match/no-input, DTMF fallback, and escalation. - Configure and test intents, entities, prompts, dialogue/workflow logic, integrations, and guardrails in our selected voice-AI platform. - Integrate with Five9 or equivalent contact-center infrastructure: SIP REFER handoff, call routing, queues/skills, call variables, transfers, and agent context. - Preserve and validate SIP headers, correlation IDs, caller/routing context, and transfer metadata across the Five9 → voice-AI → Five9 round trip. - Work with telephony engineers on SIP trunk/SBC behavior, routing, failure modes, and end-to-end test calls. - Partner with application teams on API contracts, authentication, error handling, and safe use of customer data. - Build human-handoff paths that pass the live agent the right conversation summary, caller context, detected intent, and reason for transfer. - Own test plans, UAT support, production cutover, rollback criteria, and post-launch triage. - Produce durable documentation: solution architecture, call flows, API/integration maps, runbooks, troubleshooting trees, test scripts, and knowledge-transfer sessions. - Work within Jira/ServiceNow-style delivery and ITIL change-management processes. Required experience - Shipped production voice or conversational AI agents for customer-facing contact-center use cases. - Strong practical experience with at least one modern voice-AI platform (Retell, Vapi, Bland, PolyAI, Cognigy, Google CCAI, Amazon Connect, etc.). - Experience integrating agents with a CCaaS/contact-center platform such as Five9, Genesys, NICE CXone, Talkdesk, Amazon Connect, Twilio Flex, or similar. - Hands-on understanding of **SIP call control**, especially SIP REFER, header propagation, SBC/trunk behavior, and contact-center call-routing context. - Experience with REST APIs, JSON payloads, authentication, customer-data boundaries, retries/timeouts, and failure handling. - Ability to design for real-world call behavior: barge-in, ambiguity, confirmation, retries, no input, low confidence, and clean live-agent escalation. - Comfortable leading requirements workshops with business stakeholders and working directly with application/API engineers. - Strong written documentation and knowledge-transfer habits. Nice to have - Five9 experience. - SIP REFER architecture in a Five9 or comparable CCaaS environment. - Experience with scheduling, payment, account lookup, customer support, or other transaction-oriented voice flows. - Experience operating in a regulated, change-controlled, or security-sensitive environment. - Conversational analytics, evaluation harnesses, call-transcript review, or iterative containment/transfer-rate optimization. What success looks like - A reliable voice agent handles the agreed self-service scenarios without trapping callers or creating dead ends. - Escalated calls reach the right live-agent queue with usable context. - A Five9-anchored SIP REFER flow has been tested for successful handoff, return transfer, header/context preservation, and defined failure handling. - Integrations are secure, observable, and resilient to backend failures. - The internal team receives documentation and a practical support/ownership model, not a black-box implementation. - The first release creates a reusable foundation for future voice-AI use cases.