AI-Powered Learning Management System with Edge Computing
A mobile-first LMS that runs AI inference entirely on local hardware — no cloud required.
EduSync is an edge-deployed Learning Management System built as a BTech ECE Final Year Project (12 Credits). Teachers upload study materials and the system automatically generates quizzes, summaries, and flashcards using a locally running Llama-3 model. Students take quizzes with real-time camera-based attention monitoring, while scroll engagement is quantified through DSP signal processing.
All AI processing happens on your local machine — OCR, quiz generation, attention analysis — ensuring data sovereignty and zero cloud dependency. The entire pipeline runs on consumer hardware (NVIDIA RTX 3060, 6GB VRAM).
| Feature | Description |
|---|---|
| Material Upload | Upload PDFs, images, or PPTs — AI extracts text via OCR |
| AI Content Generation | Auto-generate summaries, flashcards, and quizzes from materials |
| Classroom Management | Create classrooms and share join codes with students |
| Student Analytics | Monitor quiz scores, engagement metrics, and attention data |
| Feature | Description |
|---|---|
| Smart Study | Access AI-generated summaries and flashcards for each material |
| Quiz Assignments | Take teacher-posted quizzes with instant scoring |
| Attention Tracking | Camera-based head pose estimation monitors focus during quizzes |
| Progress Dashboard | View scores and track learning progress |
| Component | Technology | Purpose |
|---|---|---|
| Text Extraction | EasyOCR (CPU) | OCR from PDFs, images, and slides |
| Content Generation | Llama-3 8B, 4-bit quantized (GPU) | Quiz, summary, and flashcard generation |
| Attention Monitoring | MediaPipe Face Mesh + cv2.solvePnP (CPU) | Head pose estimation (yaw/pitch) during quizzes |
| Engagement Scoring | FIR filter, ZCR, FFT (CPU) | DSP-based scroll behavior analysis |
|
Frontend React Native (Expo SDK 54) Expo Router expo-camera TypeScript |
Backend FastAPI SQLite + SQLAlchemy llama-cpp-python MediaPipe + OpenCV EasyOCR |
EduSync/
├── backend/ # FastAPI backend (edge node)
│ ├── main.py # REST API endpoints
│ ├── llm_service.py # Llama-3 inference (quiz/summary/flashcards)
│ ├── parser_service.py # OCR pipeline (EasyOCR + pypdf)
│ ├── vision_service.py # Head pose attention (MediaPipe + solvePnP)
│ ├── signal_processor.py # DSP scroll engagement (FIR, ZCR, FFT)
│ ├── database.py # SQLite models & schema
│ ├── auth.py # JWT authentication
│ ├── metrics_logger.py # Research metrics collection
│ ├── research_data/ # Collected metrics & figures
│ └── scripts/ # Analysis & visualization scripts
│
├── EduSyncApp/ # Expo/React Native mobile app
│ ├── app/ # Screens (expo-router file-based routing)
│ │ ├── index.tsx # Login screen
│ │ ├── register.tsx # Registration screen
│ │ ├── intro.tsx # Onboarding video
│ │ ├── (tabs)/ # Main tab navigation
│ │ │ ├── explore.tsx # Classrooms (create/join)
│ │ │ ├── index.tsx # Materials & study
│ │ │ ├── assignments.tsx # Quiz assignments
│ │ │ └── progress.tsx # Analytics dashboard
│ │ ├── classroom/[id].tsx # Classroom detail
│ │ ├── material/[id].tsx # Material viewer
│ │ └── assignment/[id].tsx # Quiz with attention tracking
│ ├── components/
│ │ ├── AttentionTracker.tsx # Camera-based attention monitoring
│ │ ├── PrivacyConsent.tsx # Analytics consent modal
│ │ └── Skeleton.tsx # Loading skeleton
│ ├── context/
│ │ ├── AuthContext.tsx # Authentication state
│ │ └── IntroContext.tsx # Onboarding state
│ └── lib/
│ ├── api.ts # API client (axios)
│ └── config.ts # Backend URL config
│
├── docs/ # Documentation
│ ├── PROJECT_REPORT.md # Full project report
│ ├── PROJECT_OVERVIEW.md # Architecture overview
│ └── RESEARCH_PAPER_BRIEF.md # Conference paper brief
│
└── paper/ # LaTeX paper & presentation
├── edusync_ieee_paper.tex
└── edusync_presentation.tex
- Python 3.10+ with pip
- Node.js 18+ with npm
- NVIDIA GPU with CUDA (for Llama-3 inference)
- Expo Go app on your phone (iOS / Android)
# Clone the repo
git clone https://github.com/Triplejw/EduSync.git
cd EduSync/backend
# Create virtual environment
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Start the server (bind to all interfaces for mobile access)
uvicorn main:app --reload --host 0.0.0.0 --port 8000cd EduSyncApp
# Install dependencies
npm install
# Start Expo
npx expo start- Open Expo Go on your phone
- Scan the QR code from the terminal
- Configure the backend URL in
EduSyncApp/lib/config.ts:
// Use your computer's local IP address
const API_URL = 'http://192.168.x.x:8000';Tip: Run
find_backend_ip.shto find your machine's IP address on the local network.
Authentication
| Method | Endpoint | Description |
|---|---|---|
POST |
/login |
Login with email/password, returns JWT |
POST |
/register |
Register as teacher/student, returns JWT |
AI Pipeline
| Method | Endpoint | Description |
|---|---|---|
POST |
/extract-text |
OCR text extraction from uploaded file |
POST |
/generate-quiz |
Generate quiz questions from text |
POST |
/generate-summary |
Generate summary from text |
POST |
/generate-flashcards |
Generate flashcards from text |
Classrooms & Materials
| Method | Endpoint | Description |
|---|---|---|
GET/POST |
/classrooms |
List or create classrooms |
POST |
/classrooms/join |
Join classroom by code |
GET/POST |
/materials |
List or upload materials |
GET/POST |
/assignments |
List or create assignments |
POST |
/assignments/{id}/submit |
Submit quiz answers |
Analytics & Research
| Method | Endpoint | Description |
|---|---|---|
POST |
/analyze-attention |
Submit camera frame for attention analysis |
POST |
/submit-analytics |
Submit scroll engagement data |
GET |
/teacher/dashboard-stats |
Aggregate student metrics |
GET |
/research/metrics-summary |
Research metrics for report |
This project validates two research objectives for an ECE capstone and conference paper:
Objective 1 — Edge AI Performance: Demonstrate measurable inference performance (latency, throughput) for OCR and LLM operations on consumer-grade hardware, with vision-based head-pose attention monitoring.
Objective 2 — Multimodal Engagement Correlation: Demonstrate that scroll-based engagement (DSP) and vision-based attention correlate with quiz performance across users and sessions.
Research data, analysis scripts, and visualization tools are in backend/research_data/ and backend/scripts/. See the Project Report and Research Paper Brief for methodology and results.
┌─────────────────────┐ ┌─────────────────────────────────┐
│ Mobile App │ │ Edge Node (Local Server) │
│ (Expo/React │ HTTP │ │
│ Native) │◄───────►│ FastAPI Backend │
│ │ JWT │ ├── Llama-3 8B (GPU) │
│ ┌───────────────┐ │ │ ├── EasyOCR (CPU) │
│ │ Camera Feed │──┼────────►│ ├── MediaPipe Face Mesh (CPU) │
│ │ (2s interval) │ │ Base64 │ ├── DSP Signal Processor │
│ └───────────────┘ │ │ └── SQLite Database │
│ │ │ │
│ ┌───────────────┐ │ │ All AI runs locally — │
│ │ Scroll Events │──┼────────►│ no cloud dependency │
│ └───────────────┘ │ │ │
└─────────────────────┘ └─────────────────────────────────┘
EduSync uses JWT (JSON Web Tokens) for authentication. On login or registration, the backend returns an access_token and user object. The mobile app stores the token in AsyncStorage and sends Authorization: Bearer <token> on every request.
Set EDUSYNC_JWT_SECRET as an environment variable for production deployments.
Built with Edge AI for data-sovereign learning
BTech ECE Final Year Project