IMBIshimwe Mugisha Benjamin

Case study · 08 / 10

Claire AI

A voice first AI tutor that uses retrieval to keep lessons accurate.

Organisation
AI tutor project
Role
AI engineer
When
Voice tutor
Claire AI tutor interface
Claire AI, public site

Context

Learning by talking is faster for many students, but a voice tutor that guesses is worse than none.

What I built

  1. 01Integrated large language models into voice applications.
  2. 02Implemented RAG to reduce hallucinations and keep content accurate.

System view

Architecture of Claire AI, drawn as a diagram

Outcome

  • Lessons answered from course content instead of model memory.

Tools

  • Python
  • LangChain
  • RAG
  • Speech to text