LLM-Based Customer Q&A Agent

I’m ready to spin up a production-grade AI agent that can pull information from my existing database / cloud storage and answer customer questions in natural language. The core of the job is the full cycle of building and shipping the LLM features: retrieval pipeline, prompt design, orchestration layer, and robust backend endpoints.

Here is the landscape you’ll step into: the source material that the agent must reference already lives in a structured cloud bucket and a managed database. Your mission is to ingest that content, index it efficiently—vector store, embeddings, or another performant approach of your choice—and expose a clean API that lets my applications hand a user question to the model and receive a concise, cited answer back in milliseconds. Python or Node.js is fine as long as the code is readable, well-tested, and scalable; FastAPI, Express, LangChain, LlamaIndex, or similar libraries are all welcome if they help you ship fast while keeping quality high.

Deliverables
• Data-ingestion script(s) that pull from the storage layer, chunk intelligently, embed, and update the index incrementally
• Backend service with endpoints for /ask, /feedback, and /health, containerised for easy deployment
• Prompt and retrieval logic tuned for factual accuracy and low latency, with environment variables for swapping between OpenAI, Anthropic, or local models
• Clear README plus minimal tests so my team can extend the code confidently

Acceptance criteria
– Agent returns correct, referenced answers for a supplied test suite of 100 sample questions with ≥90 % accuracy
– P95 latency under 2 s for documents up to 500 k tokens
– Code passes linting and all unit tests

Please share links to similar AI or backend projects you’ve delivered, outline the stack you’d prefer using, your typical turnaround time, and your hourly or project-based rate. Strong problem-solving skills, reliable communication, and clean code matter more to me than fancy slides—let’s build something solid together.