AI Agent Architect — Voice Agent + Enterprise RAG for Fintech Rollout (LangGraph, LLM Fine-Tuning)

We're a 400-person financial-services firm rolling out AI voice agent capabilities across our CSR call center and an internal RAG chatbot for our compliance and audit teams. We're mid-pilot and need a senior AI agent architect to move us from a working prototype into a production system our ops team can trust 24/7. The scope Two parallel workstreams: AI voice agent for CSR call center — telephony integration, tool-calling for account lookup and appointment booking, real-time fallback logic. We're on LiveKit for the voice layer, with LangGraph orchestrating agent behavior. Enterprise RAG chatbot for internal compliance and audit staff — multimodal RAG over our internal compliance library (regulatory PDFs, audit guidelines, internal SOPs). Need OCR + table parsing for semi-structured docs, plus source-grounded citations on every answer. We want you to own the architecture, ship v1 to production, and stay on as a fractional technical lead through the next 3-6 months as we scale. What we're looking for Deep experience with LangGraph, LangChain, and MCP servers for tool-calling Multiple production AI agent systems (not pilots, not demos) Hands-on with LLM fine-tuning on open-source models (Llama, Qwen, Phi) Enterprise RAG deployments, ideally with multimodal / OCR AI voice agent stack: LiveKit or comparable, telephony experience Comfort operating as fractional CTO / technical advisor, not just IC Regulated-industry experience (financial services, healthcare, Big 4) is a strong plus. Please share a link to a production AI agent system you built (voice or RAG), and one paragraph on how you'd architect the compliance RAG layer. We'll shortlist based on those two things.