MathCrew

• 

  student's pin

• 

  dashboard

• 

  ss1

• 

  student name & password

• 

  motivation icons

• 

  ss5

MathCrew started as a personal project to make math practice more effective and engaging for young learners. Traditional methods rely on static worksheets and provide limited feedback, making it hard for students to understand their mistakes. The goal was to build a system that not only gives answers but also explains concepts, adapts to each student’s needs, and makes learning enjoyable using AI. What it does

MathCrew is an AI-powered adaptive math tutor for Grade 1–6 students. It generates personalized problems based on student performance, analyzes mistakes, and provides targeted feedback. It also includes gamification features like XP, levels, streaks, and badges, along with a dashboard to track progress.

How we built it

The system uses a multi-agent architecture built with CrewAI. Four agents—Manager, Creator, Helper, and Analyst—work together to generate problems, evaluate answers, analyze errors, and guide learning. Google Gemini powers the core AI capabilities. The backend is built using Python and Starlette, with SQLite for storage and SSE for real-time updates. The frontend uses JavaScript and Chart.js.

Challenges we ran into Designing a smooth multi-agent workflow Ensuring accurate mistake classification Balancing AI response quality with speed and cost Implementing real-time updates effectively Accomplishments that we're proud of Built a fully working multi-agent AI tutoring system Successfully integrated Google Gemini for adaptive learning Created real-time AI pipeline visualization Combined learning with gamification for better engagement What we learned How to design and orchestrate multi-agent AI systems Practical use of Gemini for real-world applications Importance of adaptive feedback in education Optimizing performance using caching and system design What's next for MathCrew Add voice-based interaction for younger students Expand to more subjects beyond math Improve personalization using deeper learning analytics Deploy as a scalable product for wider student access

Built With

• 2.5
• 3.10+
• backend
• chart.js
• crewai
• events
• flash
• framework)
• gemini
• google
• javascript
• local
• ollama
• python
• realtime
• server-sent
• sqlite
• starlette