Retell AI Voice Agent Architect / Developer for Retail Inbound Voice Agent

We are building a production AI voice agent for a retail-focused B2B SaaS product. The agent will answer inbound calls on behalf of our customers, identify caller intent, handle 10–12 defined use cases, call backend API functions, and route or fall back appropriately when the caller’s request is outside the supported flows. We have already ruled out the single-prompt approach. We are committed to building this in Retell using Conversation Flow. We are not looking for someone to debate whether this should be one large prompt, a single-agent prompt stack, or a general chatbot-style implementation. The architecture direction is Conversation Flow. We already have all backend APIs implemented. This project is not about designing or building our APIs. The work is focused on Retell architecture, Conversation Flow implementation, function calling design, node/component structure, prompt quality, call-flow reliability, and production readiness. At this time, we are only considering contractors who have actually shipped a complex Retell Conversation Flow or equivalent AI voice conversation-flow implementation for a paying customer with real production call volume. Prototype-only experience is not enough. We have already built versions of this agent on other voice AI platforms and moved away from them because we could not achieve the caller experience, repeatability, and reliability we need. We are now building on Retell AI and need someone with strong hands-on experience designing production-grade Retell Conversation Flows. This is not a simple chatbot prompt-writing project. We need someone who understands AI voice agent architecture, Retell conversation nodes, components, extractor nodes, function calling, latency tradeoffs, and production voice-call behavior. What We Are Building The agent will: Handle inbound retail customer service calls Support 10–12 specific use cases Allow callers to present multiple use cases in one call Use fallback handling for unsupported or unclear requests Call 1–4 already-implemented backend API functions depending on the use case Authenticate backend API calls using Retell custom headers Receive calls forwarded from Twilio to Retell via SIP Use Retell inbound webhook context to determine which business/customer the agent is representing on each call Maintain a high-quality caller experience with low latency and repeatable behavior Current Direction Based on guidance from Retell and our own testing, we are proceeding with Retell Conversation Flow. The current prototype has grown to 80+ nodes. We need help improving the architecture, reducing unnecessary complexity, and implementing best practices so the agent is maintainable, reusable, and production-ready. We are specifically looking for someone who can take an existing complex Conversation Flow direction and make it stronger, cleaner, faster, and more reliable. Retell Best Practices We Want Followed We are looking for someone who understands and can apply the following Retell design principles: Use conversation nodes for complex, multi-step, multi-scenario call flows Use components for organization and reuse where appropriate Avoid unnecessary node bloat when simpler inline extraction or function schemas are sufficient Use extractor nodes only when the extracted value affects routing, downstream logic, persistence, post-call analysis, or webhook payloads Use inline extraction in the function schema when a value is only needed for a single function call Avoid unnecessary LLM steps that could add latency Structure prompts clearly and consistently at the node level Design reusable shared functions/components across agents where appropriate Keep the caller experience fast, natural, and reliable Required Experience You must have direct experience with: Retell AI or an equivalent production AI voice conversation-flow platform Complex Conversation Flow architecture Production voice agents handling real caller volume Retell conversation nodes or equivalent node-based call-flow systems Retell components or reusable flow/module design Retell function calling or equivalent tool/function execution Extractor nodes or structured variable extraction in live voice flows Inbound webhooks and dynamic call context Backend API integration from voice-agent function calls Custom headers or equivalent API authentication patterns Twilio SIP forwarding or telephony integrations AI voice agent latency optimization Prompt engineering for voice agents Designing deterministic, repeatable conversational flows Building production customer-service or retail voice agents Strong Plus Experience with any of the following is a plus: Retail customer service workflows B2B SaaS platforms Multi-tenant agent architecture Dynamic account/business context injection Post-call analysis/webhook payload design Voice AI QA/testing frameworks Function schema design for LLM-based agents Retell deployments at production scale What We Need From You We need someone who can help us: Review our current Retell Conversation Flow prototype and architecture Recommend the right node/component/function structure Identify where extractor nodes should be used versus inline function extraction Simplify and organize the flow without reducing reliability Implement or guide implementation of the final Retell agent Improve prompt structure at the node level Ensure the agent can handle multiple caller intents in a single call Ensure calls to our already-built backend APIs are clean, authenticated, and reliable Reduce latency where possible Help us get to a production-ready implementation Deliverables Expected deliverables may include: Retell Conversation Flow architecture review Recommended node/component/function structure Updated Retell agent implementation Prompt improvements for conversation nodes Function schema recommendations Extractor node versus inline extraction recommendations API function-call integration review Testing plan for supported use cases and fallback cases Documentation of the final architecture Ideal Candidate The ideal candidate has shipped real AI voice agents into production, not just built demos, prototypes, or text chatbots. You understand that voice calls require fast responses, clear routing, reliable function calls, graceful handling of ambiguity, and measurable consistency across real caller traffic. You should be able to explain tradeoffs clearly and make practical implementation decisions based on production experience. Not a Fit Please do not apply if: You have only built chatbot-style prompts You have only built simple single-prompt voice agents You have not shipped a complex conversation-flow voice agent for a paying customer You do not have experience with production call volume You want to persuade us to use a single-prompt architecture instead You are primarily an API/backend developer without deep voice-agent flow experience