Skip to content
William Peytz

2025 · BSc thesis · full-stack · ML

EloQuiz.dk

Adaptive math learning platform that uses an Elo-based difficulty system to match each student to the right next question.

  • Vue
  • Flask
  • Firebase
  • OpenAI API

Overview

EloQuiz.dk is an adaptive math learning platform built around an Elo-rating system: every student and every question has a rating, and the matchmaking loop tries to keep students at the productive edge of their ability.

This started as my BSc thesis on adaptive learning systems and grew into a real product validated through a student pilot.

What I built

  • A full-stack web app (Vue front-end, Flask back-end, Firebase for auth and data) deployed end-to-end.
  • The Elo-based difficulty engine: students and items co-rate each other, letting the system surface problems in a student’s zone of proximal development without manual curriculum tagging.
  • LLM-assisted question generation and explanation flows powered by the OpenAI API.

What I learned

How brittle a “smart” model gets the moment real students start using it. Most of the engineering work was the loop around the model: onboarding, explanations, error states, and trust. The Elo math was the easy part.