A RAG-powered assistant over a course knowledge base

Retrieval-augmented generation grounding student support answers in verified course material.

Close-up of networking cables in a data center

The challenge

Student support was fielding the same course-content questions repeatedly, and existing chatbot attempts hallucinated confident, incorrect answers.

Our approach

We built a retrieval-augmented generation system indexing course materials, FAQs, and policy documents, with source citations on every answer and an evaluation harness to track grounding accuracy over time.

Results

  • Majority of routine student questions handled without a human
  • Every answer traceable back to a cited source document
  • Ongoing evaluation pipeline catching regressions before release
Start A Project