Agentic RAG System with Vector Database for Large Files and Books Long Term AI Product

## JOB DESCRIPTION I am looking for an experienced AI engineer ( that have already worked on this type of project in the past ) to help design and build an Agentic RAG system with a vector database The core idea is to efficiently process and query large documents such as books large PDFs technical manuals and knowledge bases using chat based interaction This is not a short task The goal is to build a scalable production ready AI system that can evolve into a full product with long term collaboration --- ## CORE IDEA AND GOAL The system must -- ingest large files including books and long PDFs -- chunk and index them efficiently into a vector database -- allow high quality conversational querying over large knowledge sources -- maintain performance accuracy and low cost even at scale Efficiency and scalability are critical from day one --- ## TECHNICAL CONTEXT I currently have a working n8n workflow and I am considering using n8n as the backend However I need expert guidance on -- proper architecture and configuration -- whether n8n is suitable long term -- alternative architectures that are more flexible scalable or cost efficient I am open to better technical suggestions if justified --- ## WHAT I NEED HELP WITH - designing an Agentic RAG architecture -- choosing correct chunking strategies for very large documents -- optimizing retrieval accuracy and latency -- embedding strategies and vector database tuning -- efficient metadata labeling for documents and sections -- handling large file ingestion without performance degradation -- cost optimization for embeddings LLM usage and infrastructure -- retrieval quality measurable with evaluation examples --- ## PRODUCT AND BUSINESS QUESTIONS I also need strategic guidance beyond coding - - best way to offer this as a product to customers -- public web application versus private access per customer -- multi tenant versus single tenant architecture -- whether each customer should have their own Supabase instance or shared infrastructure -- n8n licensing and policy constraints including enterprise considerations -- data privacy isolation and compliance best practices -- secure access management for paid customers Access in my opinion should be sold individually to each customer and privacy is critical --- ## PROJECT DURATION This is a long term project - - initial MVP build -- multiple future milestones -- continuous improvements -- maintenance optimization and scaling If the collaboration works there will be ongoing work --- ## BUDGET AND COMPENSATION There is no fixed budget Compensation structure -- setup fee and paid implementation time -- long term collaboration -- percentage from recurring revenue open for discussion The agreement will depend on the MVP scope and architecture decisions --- ## APPLICATION REQUIREMENTS STRICT To apply you must -- send a Loom video explaining -- your technical approach -- architecture decisions -- how you would build and scale this system — show some examples for past projects (must have created something similar in the past) Direct messages or text only applications will be ignored This is a long term opportunity requiring real expertise and creative thinking