Work / 02
A knowledge system that answers from your own documents
An organization had a large body of documents where the answers already existed — the problem was finding them. We built an AI application that finds the right passage and answers with its source.
Your documents
Knowledge system
Employees
Every answer comes with its source, so it can be checked in one click.
Starting point
The collection of documents was large and kept growing. General AI tools gave well-written answers, but without any connection to the organization’s own material — and sometimes wrong. What was needed was something that started from their own documents and showed where each answer came from.
What we did
We built the application from scratch: a Python backend, a React frontend, document processing, vector search in Elasticsearch, and a RAG layer. When someone asks a question, the most relevant passages are retrieved and the assistant answers from them, with a reference to the source. An automated flow ingests new and changed documents.
Result
People ask questions in their own words and get answers based on the organization’s documents — with the source one click away, so every answer can be checked.