GenAiHackathon

🌍 Sustainable Agentic Tutor — AI for Equity, Education & the Planet

Empowering learning through self-hosted LLMs. Designed for education. Built for sustainability. Driven by DEI.

---

📚 Overview

This project is a next-generation educational agentic system that:

• ✅ Creates personalized learning paths using students' quiz responses
• 🔍 Uses a RAG pipeline powered by self-hosted LLMs to pull knowledge from curriculum PDFs
• 🔗 Fetches real, trustworthy learning resources (like Khan Academy & YouTube) using live search
• 🌐 Offers multi-language support with translation capabilities (e.g., English and other languages)
• 💻 Includes a Next.js web app frontend with a rich, interactive experience
• 📊 Provides a live dashboard showing energy usage, carbon savings, and cost comparisons

---

🌱 Why Sustainability + DEI?

Traditional cloud-based LLMs contribute to high energy consumption and carbon emissions. By self-hosting open-source models like Ollama, Gemma, and StableLM, this project:

• ⚡ Reduces compute footprint
• 🌎 Promotes green AI adoption
• 💰 Demonstrates real-time cost savings over services like OpenAI
• 🤝 Ensures equity by offering free and accessible education tools, adaptable across languages and learning needs

---

🧠 Agentic Workflow Summary

1. Student takes a quiz (in English or a variety of other languages)
2. PDF curriculum is uploaded
3. Self-hosted LLMs:
   • Identify student strengths and weaknesses
   • Use a RAG pipeline to generate a personalized learning path
4. Topics are extracted from the path
5. Search APIs fetch real-time educational content (Khan Academy, YouTube)
6. The LLM filters and cleans the resources for relevance

---

💻 Frontend: Built with Next.js

Our responsive web app is built using Next.js, providing a seamless user experience where students can:

• 📋 Take subject-specific quizzes
• 📎 Upload their curriculum PDFs
• 📊 View a real-time dashboard of learning recommendations and environmental insights

---

📊 Sustainability Dashboard

In the Dashboard tab of our app, we highlight measurable impact:

Metric	Description
⚡ Energy Usage	Tracks approximate energy consumption per session
🌱 Carbon Savings	Compares emissions from local vs cloud LLM usage
💸 Cost Savings	Token-level comparison: self-hosted vs OpenAI API calls

This level of transparency helps build awareness about the environmental impact of AI — and proves that sustainable AI is possible.

---

🛠️ Tech Stack

Component	Description
🧠 LLMs	Ollama (mathstral, stablelm2, gemma, deepseek)
📄 PDF Parsing	PyPDFLoader
🧩 RAG Pipeline	Chroma, LangChain, ChatPromptTemplate
🌐 Search	SerpAPI for real-time resource fetching
🗂️ Embeddings	nomic-embed-text via Ollama
🖥 Frontend	Next.js + TailwindCSS
📦 API Server	FastAPI
🧪 Caching	SQLiteCache (LangChain)

---

🚀 Local Setup

# Clone repo
git clone https://github.com/AhmedJader/GenAiHackathon

# Set up .env file
SERPAPI_API_KEY=your_key_here

# Change into correct backend directory
cd backend/src

# Install dependencies
pip install -r requirements.txt


# Run FastAPI backend
# (make sure you are in the correct working directory and are using a Python virtual environment!)
uvicorn main:app --reload


# Change into correct frontend directory
cd ../../frontend

# Install frontend dependencies
npm install

# Run Next.js frontend
npm run dev

#You're done :)!

---

💡 Future Plans

• ✅ Add voice support for accessibility
• 📈 Integrate student progress tracking
• 🧩 Plugin system for STEM subjects
• 🧠 Hugging Face-compatible deployment options