With auto-assessment, questions can be adapted to students.
Derivativ features an interactive practice area for students to work on.
Teachers can easily generate IGCSE-compliant materials with Derivativ.

Derivativ — AI-Powered Exam Prep for Math

🚀 Inspiration

Derivativ is a full-circle project for us. Ernie once taught IB tuition she couldn’t afford herself. Ali, an IGCSE student, knew the value of personalized learning. This is our gift to students and teachers everywhere.

But it’s also a response to a global problem: One in three 15-year-olds in the US lack basic math skills, according to OECD studies. In an AI-driven world, math isn’t just about passing exams — it’s the foundation for understanding the systems shaping our future.

Derivativ aims to make high-quality math preparation accessible.

✨ Try It Out

👩‍🎓 Students

👉 Get Started

👨‍🏫 Teachers

Want to generate exam materials in seconds?

👉 Try the Teacher Tool

🎯 What It Does

Derivativ uses curriculum-constrained AI to help teachers and students prepare for standardized math exams, starting with IGCSE.

For Teachers: Generate complete, exam-compliant worksheets with marking schemes in seconds — cutting content creation time by 95%.
For Students: Personalized, interactive practice with instant feedback and a drawing canvas for working out problems.

At the core is our intelligent autoassessment system, which adapts to each student’s pace in real-time. It keeps practice in the optimal learning zone — not too hard, not too easy — using performance-driven difficulty adjustment and topic-level analytics to target weak spots efficiently.

🛠 How We Built It

A full-stack, production-ready platform built on Bolt.new:

Backend: FastAPI (Python) with a multi-agent AI system — QuestionGenerator, Marker, Reviewer, Refiner — working together to create, verify, and refine every question.
Frontend: React + TypeScript for a clean, intuitive experience.
Storage: Supabase (PostgreSQL), Cloudflare R2.
AI: Multi-provider LLMs (OpenAI, Anthropic, Gemini).
Testing: 97.4% backend coverage with Pytest.

⚡ Challenges

Remote Collaboration: We are located in different timezones and synchronizing our efforts across multiple repositories and platforms has been tricky!
Guaranteeing Accuracy: Solved with our multi-agent review loop and structured outputs to ensure syllabus alignment.
Production-Grade Quality: Reliable, scalable, with automated testing baked in.

🏆 Accomplishments

✔️ Curriculum-constrained AI.
✔️ Fully production-grade with high test coverage.
✔️ Sophisticated multi-agent AI system ensuring accuracy.
✔️ One seamless platform serving both teachers and students.

💡 What We Learned

In education, constraint is a feature. AI only works for exam prep when strictly limited to the curriculum — that’s where it becomes truly effective.