School Agent – Project Story

📌 About the Project

When exams arrived, my classmates and I struggled to locate quick answers from our syllabus. We spent plenty of time hunting through PDFs and textbooks rather than learning.

Concurrently, I was particularly fascinated by AI models and fine-tuning, so naturally, I was wondering how I could fix this issue with tech. That is when I opted to create School Agent, an AI-fueled Retrieval-Augmented Generation (RAG) platform that assists learners in immediately finding answers within their syllabus.

🛠️ How I Built It

I built School Agent with:

Python for backend logic
Streamlit for the frontend and UI, making it convenient to interact with the chatbot
Ollama to execute AI models on local machines
FAISS as the vector database to store and fetch syllabus data efficiently
SQLite to store user data

How It Works

Admin uploads syllabus files 📄
The AI processes the documents and saves embeddings in FAISS
Students submit questions 💬
The chatbot fetches correct answers from the syllabus using RAG

With Streamlit covering both UI and logic, I did not require an external API, which simplified the deployment.

🚀 Challenges & What I Learned

Creating School Agent was a thrilling challenge! Some major takeaways:

✅ Managing Big Syllabus Files – Fine-tuning FAISS for quick and precise retrieval
✅ Enhancing Response Precision – Fine-tuning embeddings and improving the way the chatbot interprets queries
✅ Streamlit UI for Easy Use – Building an intuitive interface without frontend frameworks
✅ Security & Access Control – Making sure only admins could upload syllabus files while students had limited access