Description
Summary We run a growing beauty and aesthetics business (treatments, bookings, consultations) and most of our client communication is still manual — answering the same questions over WhatsApp/Instagram/email, chasing no-shows, and manually deciding what needs a human vs. what’s routine. We want to move from “someone typing replies all day” to a real agentic system that can think, route, and act. What we need: We’re looking for someone who can design and build a multi-agent workflow that: • Classifies incoming client messages (booking requests, treatment questions, complaints, general inquiries) by intent and urgency • Uses RAG over our treatment menu, pricing, aftercare instructions, and FAQs to answer client questions accurately and consistently • Escalates sensitive or complex cases (medical concerns, complaints, high-value bookings) to a human with a clean summarized handoff • Integrates with our booking/CRM system and messaging channels (WhatsApp, Instagram, email) via API/webhooks • Is observable — we want to see what the agents are doing and why, not a black box Must-have: • Production experience with agent orchestration frameworks (LangGraph, CrewAI, or similar) — not just tutorials • Strong RAG implementation experience (chunking, retrieval quality, grounding/citations) • Comfortable building end-to-end: prompt design, orchestration logic, API integrations • Clear communicator, can explain technical tradeoffs to a non-technical team Nice-to-have: • Experience in beauty, aesthetics, or another appointment-based service business • Experience connecting agent systems to no-code/low-code automation layers (n8n, Make, Zapier) • Familiarity with booking/CRM platforms • Past work automating a real client-facing workflow, with before/after impact (time saved, missed bookings recovered, etc.) How to apply: Share one multi-agent or RAG system you’ve built in production, what problem it solved, and the measurable outcome. Also tell us your default stack for a project like this and why.