Build a Multi-Tenant AI Support Copilot SaaS (RAG + AI Agent + Next.js + FastAPI + n8n)

We're looking for an experienced full-stack and GenAI engineer to build the MVP of a multi-tenant SaaS. Businesses will use it to turn their internal documents into an AI support assistant. This is not a basic chatbot wrapper. We need a production-minded system with: - an async document ingestion pipeline - hybrid retrieval with citations - an agent that can call tools and escalate to humans - an admin dashboard with analytics - an embeddable chat widget for customer websites You should be comfortable working across a TypeScript frontend, a Node.js API, a Python AI service, PostgreSQL with pgvector, Redis queues and n8n automation. Clean architecture, sensible trade-offs and good documentation matter as much as working features. What we want to build: - Tenant workspace: Organizations sign up, invite team members (Owner/Admin/Agent roles), and upload PDFs, DOCX files, or website URLs as their knowledge base. - Ingestion pipeline: Uploaded content is processed in the background (parse, chunk, embed, index), with live status updates in the UI. - AI Support Agent: This is a streaming chat assistant that: - answers from the tenant's documents with source citations - routes each question by intent - uses tools such as creating a support ticket - hands off to a human when it isn't confident - Embeddable widget: A single script tag lets a tenant add the assistant to their own website. - Admin dashboard: Shows conversations, low-confidence answers, escalations, deflection rate and document management. - Automation: Escalations trigger an n8n workflow that notifies the team via Slack or email. - Privacy layer: PII is redacted before anything reaches the LLM, and an audit log records sensitive actions. Tech stack: Next.js (TypeScript), Node.js (NestJS), Python (FastAPI), LangChain or LlamaIndex, PostgreSQL + pgvector, Redis (BullMQ), n8n, Docker Compose. Duration: 2 weeks Budget: $1,500 fixed price, split into 3 milestones To apply, please include: 1. A short description of a RAG or AI agent system you've shipped to production. 2. How you would combine vector search and keyword search for better retrieval. 3. Your availability over the next 2 weeks. A detailed technical specification will be shared with shortlisted candidates.