AI-Powered RAG Chatbot for Internal Knowledge Base (React + Python + LangChain)
Budget: $30 – $250 USD
I need a full-stack engineer who can take roughly 200 pages of internal policy and product PDFs and turn them into a smooth retrieval-augmented-generation experience for our team. Your first job will be to ingest those PDFs, chunk them sensibly, and store the embeddings in a vector database—pgvector is my default choice, though I am open to any alternative you can justify.
On top of that store you will wire up a RAG flow in LangChain, leaning on its Integration with OpenAI or Claude so answers come back in natural language and include transparent source citations. A lightweight FastAPI backend should expose a single chat endpoint that handles retrieval and generation, while a simple, functional React front end lets users ask questions and see cited answers. No fancy UI polish needed; speed and clarity matter more.
We run everything on AWS today, so you will package and deploy the service there—EC2, Lambda, or another setup you feel is right for the load profile. Please outline your reasoning when you propose the architecture.
To succeed you must already have production experience with RAG pipelines, vector databases, Python, React, LangChain (or a comparable orchestration framework), and AWS deployments.
Deliverables
• A working chatbot reachable via the React interface, returning answers with clickable source citations
• Clean, developer-oriented documentation covering setup, deployment, and ongoing maintenance
If this sounds like your wheelhouse, tell me how you would approach data ingestion, vector search, and scaling, and include links or brief notes on similar projects you have shipped.
On top of that store you will wire up a RAG flow in LangChain, leaning on its Integration with OpenAI or Claude so answers come back in natural language and include transparent source citations. A lightweight FastAPI backend should expose a single chat endpoint that handles retrieval and generation, while a simple, functional React front end lets users ask questions and see cited answers. No fancy UI polish needed; speed and clarity matter more.
We run everything on AWS today, so you will package and deploy the service there—EC2, Lambda, or another setup you feel is right for the load profile. Please outline your reasoning when you propose the architecture.
To succeed you must already have production experience with RAG pipelines, vector databases, Python, React, LangChain (or a comparable orchestration framework), and AWS deployments.
Deliverables
• A working chatbot reachable via the React interface, returning answers with clickable source citations
• Clean, developer-oriented documentation covering setup, deployment, and ongoing maintenance
If this sounds like your wheelhouse, tell me how you would approach data ingestion, vector search, and scaling, and include links or brief notes on similar projects you have shipped.