AIFeatured project
NeuralDesk
AI Support Agent
−64% first-response time
The challenge
Support teams were answering the same questions from scattered docs and old tickets, and first responses were slow.
What I built
A RAG-powered customer-support assistant that ingests docs & tickets into a vector store and answers with cited sources, cutting first-response time dramatically while keeping a human in the loop.
- 01Ingested docs and past tickets into a vector store with source metadata on every chunk.
- 02Answers are generated only from retrieved passages and cite their sources.
- 03Kept a human in the loop: agents review and send, and their edits improve future answers.
The result
−64% first-response time