Project — data-battle-ia
An AI-Driven Platform to Generate Law Exam Questions
Semi-finalist – 2-week AI competition.
In plain English
This project aimed to help law students study more efficiently. The platform analyzed hundreds of documents. Including lessons, exams, and corrections and automatically generated multiple-choice questions to help students train.
For the curious
The system relied on a RAG architecture: documents were manually cleaned, vectorized sentence-by-sentence, and embedded for semantic search. A Python + FastAPI backend handled question and answer generation, while a Spring Boot API managed user accounts, JWT authentication, and data storage. The front-end was built with SvelteKit, and PostgreSQL stored all generated QCMs. My role included dataset cleaning, embedding generation, authentication, and developing various backend services and controllers.
A little summary :
This project combined NLP, RAG systems, multi-backend architecture, and user-facing educational tools. Reaching the semi-finals validated the robustness and creativity of the approach.