Final Year Project · UTAR · 2025–2026

AI Academic Assistant

A full-stack AI academic platform built across 4 repositories — cross-platform frontends, a production Supabase backend, and a Python AI microservice with 6 specialized LangGraph agents for adaptive learning, autonomous scheduling, and semantic knowledge processing.

Next.jsReact NativeFastAPILangGraphSupabaseOR-ToolsPythonTypeScriptGemini API

fyp-ai-service ↗ Live Frontend ↗ Request Demo →

Demo

Demo video coming soon.

System Overview

The platform operates across 4 tiers:

Client Tier: Next.js web + React Native (Expo) mobile
BaaS Tier: Supabase (PostgreSQL, Auth, Edge Functions, Realtime)
AI Service Tier: FastAPI + 6 LangGraph agents
External: Google Gemini API + Mock LMS

Security model

Frontends never call the AI service directly. All AI requests traverse Supabase Edge Functions with Bearer token authentication — the AI service URL is never exposed to the client.

Infrastructure Highlights

Security

RLS on all 19 tables
Edge Function gateway
Signed URL storage
Bearer auth gateway

Adaptive Learning

FSRS + BKT-weighted EMA mastery scoring
FSRS review scheduling per concept

Background Infra

3 asyncio loops
No Celery, no Cloud Scheduler
Zero external dependencies

Schedule a Demo

The AI service runs a persistent asyncio background scheduler and cannot be hosted on serverless platforms.

Schedule a Demo Session →

I'll spin up a session — Parser Agent, RAG quiz generation, and CP-SAT scheduling available to demo live.

Back to Portfolio