AI Learning Platform — 2026
A learning platform that turns study material into quizzes with daily challenges, XP, and streaks. Backend infers student comprehension from assessment results.
- Role
- AI Learning Platform
- Services
- ReactTypeScriptFastAPIPython
- Year
- 2026
Challenge
Study material is passive. Asahlagi exists to make it active: turn any notes into spaced daily quizzes, then keep momentum with XP and streaks. The part that matters is the inference behind it, understanding what a student actually gets wrong so the next set is aimed at the gap.
Approach
A React and TypeScript frontend feeds a FastAPI backend, with a 52-test suite guarding the logic. The scoring path turns assessment results into a comprehension signal per topic, which drives which questions surface next. The question generation itself sits on the IndoT5 model layer, fine-tuned for Indonesian and served through a Hugging Face Space.
Outcome
Winner of Best Capstone Project on the Dicoding AI learning path. The full path is mine end to end: the model on the Hub, the inference Space, the FastAPI backend, and the React interface on top.