Joyce

PORTFOLIO

Back to Projects
Completed2026· Built with Ramya Sri Muthuluri

NexusEdu

Local Build · No public repo

ABOUT THIS PROJECT

Hackathon-built AI tutoring system with multi-agent workflows that adapt learning paths by grade level and style.

Built as part of a group hackathon for our MSc course module. NexusEdu uses n8n multi-agent workflows to generate personalised learning paths, with OpenRouter routing to Llama for LLM-powered content generation tailored by grade level. Supabase handles persistent student data and session memory. Self-hosted, no public deployment.

WHAT I LEARNED

Multi-agent n8n workflows require careful handoff design — each agent needs clean input/output contracts to avoid cascading failures

OpenRouter makes it easy to swap LLMs without rewriting pipeline logic — huge advantage in a hackathon where you're iterating fast

Personalising content by grade level is harder than it sounds — the prompt engineering to hit the right reading complexity took most of our time

Supabase is genuinely fast to set up for session memory — we had persistent student profiles working within an hour

Hackathons teach you that a working demo with one feature beats a broken demo with ten

TECH STACK

n8nOpenRouterLlamaSupabase
Back to all projects