Applied AI & Agents · 2026Hackathon winner

Synapse AI

Multi-agent tutoring platform that turns uploaded materials into personalized study outputs with teacher oversight.

AI AgentsMCPFastAPISupabaseNext.jsHITL

Problem

Students need personalized study support, but teachers also need visibility into diagnostics, progress, and the learning materials being generated.

Approach

Our team of three combined a multi-agent backend, an MCP server for course material, FastAPI, and Next.js to generate notes, flashcards, quizzes, podcasts, and streamed tutoring from uploaded course materials.

Result

Synapse AI won the AgentShyft Hackathon in May 2026.

How it works

Step 1 of 6: Upload learning material

Highlights

  • Won the AgentShyft Hackathon with a multi-agent tutoring platform: agents call course-material tools through an MCP server and stream answers over SSE.
  • Built the teacher side in FastAPI and Supabase: classroom invites, material uploads, and class APIs.
  • The platform turns uploaded learning materials into personalized study outputs such as notes, flashcards, quizzes, and podcasts (Next.js, ElevenLabs API).

My contribution

I built the teacher side of the backend in FastAPI and Supabase: classroom invites, material uploads, class APIs, and the database migration behind them. I also added the endpoint that picks a student's weakest topics and starts the diagnostic agent. My teammates built the agent runtime and the frontend.

Why the workflow mattered

The system was designed around teacher orchestration rather than a standalone chatbot. Generated learning material stays connected to source content, student needs, and teacher oversight.

Recognition

Winner, AgentShyft Hackathon, May 2026.

Back to projects
Role Fit Check