AI Engineer Needed to Build a Production-Ready RAG Assistant with LangGraph & FastAPI

Project Overview: We are developing an AI-powered knowledge assistant for our SaaS platform and need an experienced Python/AI engineer to build the core intelligence layer. Our customers frequently ask questions already answered in our documentation, but they also need information from their own accounts and uploaded documents. We want an AI assistant that can retrieve accurate information, provide source-backed responses, securely interact with our existing APIs, and escalate questions when it cannot provide a reliable answer. The solution needs to be scalable, secure, and maintainable. We are not looking for a simple chatbot or an OpenAI API wrapper. Scope of Work ● Design and implement a RAG pipeline for PDFs, knowledge-base articles, and structured business data. ● Build document ingestion, chunking, embedding, indexing, and retrieval workflows. ● Implement semantic and hybrid search using PostgreSQL/pgvector or Qdrant. ● Develop Python/FastAPI endpoints for document processing, conversations, and AI tool execution. ● Build LangGraph workflows for query routing, retrieval, tool calling, and human escalation. ● Integrate OpenAI or Claude with structured outputs and conversation history. ● Connect the assistant securely to customer-specific account data through existing APIs. ● Implement source citations, tenant-level data isolation, and prompt-injection safeguards. ● Evaluate answer accuracy, retrieval relevance, latency, and token consumption. ● Deploy using Docker and cloud infrastructure. Required Skills ● Python, FastAPI ● LangGraph, LangChain ● Retrieval-Augmented Generation (RAG) ● OpenAI / Anthropic API ● PostgreSQL, pgvector / Qdrant ● Embeddings, Vector Search, Hybrid Retrieval ● REST APIs, Docker, AWS ● LLM Testing, Evaluation, and Observability Expected Deliverables 1. Working document ingestion and vector search pipeline. 2. LangGraph-based AI assistant with API tool integrations. 3. Secure, customer-specific document and account retrieval. 4. Source-grounded answers and human escalation. 5. Automated evaluation and backend tests. 6. Deployment instructions and API documentation. To Apply Please share a production AI application you have built using RAG, LangGraph, or FastAPI. We would also like to know: 1. How would you prevent the assistant from accessing another customer's data? 2. What approach would you use to evaluate retrieval accuracy and reduce hallucinations? 3. Have you deployed AI agents with tool calling or multi-step workflows in production? We prefer developers who can demonstrate real implementation experience and explain their technical decisions clearly.