AI customer-support chatbot

A startup

An LLM-powered support assistant, built with a startup's engineering team. I built the document ingestion pipeline, the vector search, and the chat interface.

Team: Startup engineering team
Built with: Python and LangChain with the OpenAI API, pgvector for search, and a React chat interface.
Result: The assistant now handles about a third of routine support tickets.

How it fits together

Support documents ingested and chunked
Vector search pgvector
Answer LangChain, OpenAI API
Chat interface React
Routine tickets about a third handled

What I did

I joined a startup’s engineering team to build their support assistant. My parts were the pipeline that ingests their support documents, the vector search that finds the right passages for a question, and the chat interface customers use.

Result

The assistant now handles about a third of routine tickets.