Project 05, AI and education
StudySetu
A prototype that finds what each student missed after a lecture, and builds a short revision session around it.

- Event
- TetraTHON 2026, Indo-French AI Innovation Sprint
- Stage
- Selection-round project
- Status
- Working prototype
- Users
- Students, teachers, admins
01, The idea
The teacher teaches. StudySetu checks that it stuck.
StudySetu is a revision companion, not a replacement for the teacher. Teachers build the curriculum, and their own material grounds everything the system produces.
AI writes the questions and lessons, and no AI-generated item reaches a student until a teacher approves it.
02, How it works
Diagnose first, then revise.
- [ Step 1 ]
Diagnose
The first time a student opens a topic, a five-question probe checks what they already know.
- [ Step 2 ]
Track mastery
Bayesian Knowledge Tracing, a simple statistical model of whether a skill is learned, updates after every answer.
- [ Step 3 ]
Revise
A short session is built from what the probe found, with a doubt chat grounded in the lesson itself.
- [ Step 4 ]
Review
Teachers approve AI content and see which misconceptions are common across the class.
03, Screens
Built for three kinds of users.


04, Engineering
How it is built.
- One AI gateway. Every AI call goes through one gateway with automatic failover across four providers.
- Safe deployments. Docker Compose and a GitHub Actions pipeline that rolls back a failed deployment.
- Prototype stage. It was built for a hackathon round and is not a finished product.
Built with
- React 18
- TypeScript
- Vite
- FastAPI
- PostgreSQL
- pgvector
- Docker
- GitHub Actions