Add AI document assistant to our Django + Node.js platform with Next.js front end

We run a saas platform with multiple customers. The core is Django and DRF on PostgreSQL with Celery and Redis. Two years ago we added a Node.js/TypeScript layer (Express, BullMQ) that handles real-time notifications, third-party integrations and the API our Next.js front end talks to. AWS throughout. We want an assistant that answers questions over each customer's uploaded documents and drafts standard letters from them. We have a demo in a notebook; we need it built into the product properly, across both backends. Scope: - Django app for ingestion, embeddings and retrieval: Celery pipeline from S3, pgvector on our existing Postgres, strict per-customer isolation through our existing tenant model - Node.js service for the chat layer: streaming responses to the front end, conversation state, rate limiting per customer, calling the Django retrieval API. This is where our real-time infrastructure already lives - Chat and draft-review UI in our Next.js app matching our design system - Draft generation with a review-and-edit step before anything is saved - Token budgets per customer, basic evals in CI so prompt changes don't silently get worse - Deployed on our AWS (ECS), Django admin views for ops, docs for our two engineers You'd own the whole feature across Django, Node and Next.js, and can bring one more engineer, but you're the single point of contact. Must have: Django and DRF in production at scale, Node.js/TypeScript backend in production (not just Next.js API routes), PostgreSQL, one LLM feature shipped to real users. Nice to have: pgvector, LlamaIndex or LangChain, BullMQ, MongoDB (one legacy module), regulated data.