Amine EL MAHI
Lire en français
All projects

Internship · Web4Jobs

NextCourse improvements

An AI-powered search engine for learning resources — rebuilt scraper, semantic layer and recommendations grounded in the catalogue.

What I did

  1. Rebuilt the scraper to be safer (robots.txt, rate limits) and richer: instead of just the title and description, it extracts each course’s full metadata (syllabus, prerequisites, level, language, duration…).

  2. AI layer: each course is condensed into a document and turned into multilingual embeddings; the assistant (Gemini) finds the courses semantically closest to the request, then reads their details to recommend only courses from the catalogue.

  3. Improved the data layer (Prisma) and the interface, and added email-verified authentication with roles.

  4. Containerised the application (Docker Compose) for reproducible deployments.