AI Learning

A final-year project (PFA) I designed and built entirely on my own at EMSI. Traditional e-learning platforms rely on multiple-choice quizzes and predefined answers, with no detailed pedagogical feedback. I designed a platform where AI evaluates open-ended answers and generates teaching content.

• Offer a structured learning path: courses, modules, lessons, exercises.
• Automatically evaluate students' free-text answers using the OpenAI API.
• Generate teaching content with AI on demand.
• Provide complete progress tracking: scores, advancement, statistics.
• Clearly separate student and instructor roles.

• Full-stack design and development of the platform, end to end.
• OpenAI API integration for answer evaluation and content generation.
• Development of the student and instructor journeys: courses, lessons, exercises, progress.
• Data modelling and authentication with Supabase / PostgreSQL.

• Sign-up and sign-in via email or Google account, with student or instructor role.
• Course catalog by level: beginner, intermediate, advanced.
• Courses structured into modules and lessons with objectives and key concepts.
• Practice exercises with free-text answers, automatically graded by the AI.
• AI-generated teaching content.
• Progress tracking: scores and statistics for students, overview for instructors.

• Three-tier full-stack architecture: presentation, application, data.
• React frontend with Tailwind CSS.
• Node.js / Express backend exposing the application API.
• PostgreSQL database via Supabase: authentication, roles and progress data.
• AI layer: OpenAI API calls for answer evaluation and content generation.

• React
• Node.js
• Express
• Supabase
• PostgreSQL
• Tailwind CSS
• OpenAI API

• Grading open-ended answers: crafting reliable evaluation prompts where a binary quiz is not enough.
• Role separation: student and instructor journeys and permissions differ on almost every page.
• Tracking consistency: aggregating scores and progress across courses, lessons and exercises.

• Working platform presented and defended: complete journey from sign-up to the AI-graded exercise (real screenshots below, taken from the project report).
• Known limitation: operation depends on the OpenAI API (key required) and a Supabase project — not publicly deployed.
• Known limitation: academic demonstration project with sample course content.