🧠 AI Study Assistant

An AI-powered study companion that helps students understand lecture material through intelligent question answering, slide summarization, PDF summaries, and flashcard generation. Built with LangChain, Hugging Face Transformers, and Gradio — and fully powered by open-source LLMs running on your local GPU.

---

🚀 Features

📄 Upload Lecture PDFs

Drag and drop any academic or lecture PDF — slides or notes.

❓ Ask Questions from Slides

Use natural language to ask:

• “Summarize Slide 4”
• “What is Linear Regression?”
• “What is covered under Artificial Neural Networks?”

Supports fuzzy slide matching and page number detection.

🧠 Full PDF Summarizer

Click one button to get a concise overview of the entire document.

📝 Flashcard Generator

Automatically generates Q&A-style flashcards from your material — perfect for revision and quizzes.

🧩 Slide Number Matching (Fuzzy Logic)

Queries like "Summarize Slide 5" are matched against actual page content using fuzzy logic for accurate results.

---

---

🛠 Tech Stack

Layer	Tools Used
Frontend	Gradio (Python UI)
LLM	Hugging Face Transformers (Phi-2 / Mistral 7B / Falcon)
Framework	LangChain
Vector DB	FAISS
PDF Parsing	PyMuPDF (via LangChain loaders)
Embeddings	SentenceTransformers (MiniLM-L6-v2)
Fuzzy Matching	RapidFuzz

---

📁 Project Structure

ai_tutor/
├── app.py
├── ui/
│   └── interface.py
├── modules/
│   ├── document_loader.py
│   ├── vector_store.py
│   ├── rag_pipeline.py
│   ├── llm_interface.py
│   ├── slide_mapper.py
│   ├── summarizer.py
│   └── flashcards.py
├── data/
│   └── (your PDFs here)
├── requirements.txt
└── README.md

---

⚙️ Setup Instructions

1. Clone the repository

git clone https://github.com/your-username/ai-study-assistant.git
cd ai-study-assistant

2. Create a virtual environment

python -m venv venv
source venv/bin/activate   # or venv\Scripts\activate on Windows

3. Install dependencies

pip install -r requirements.txt

4. Run the app

python -m ui.interface

---

💡 Example Questions

• “Summarize Slide 3”
• “What is supervised learning?”
• “What is covered on page 10?”
• “Generate flashcards for the regression topic”
• “Give a summary of this entire PDF”

---

📷 Screenshots

You can replace these with your actual screenshots.

---

📦 Models Tested

• microsoft/phi-2 ✅ (recommended for speed and accuracy)
• mistralai/Mistral-7B-Instruct-v0.1 ✅
• tiiuae/falcon-rw-1b ✅ (lightweight)

You can switch models in llm_interface.py

---

🔐 Local & Private

This project runs entirely offline on your own PC — no API keys or cloud calls required. Perfect for privacy and local LLM experimentation.

---

🧠 Skills Used

• LangChain chains and document loaders
• RAG (Retrieval-Augmented Generation) pipeline
• LLM fine-tuned prompting
• Embedding & similarity search with FAISS
• PDF text extraction using PyMuPDF
• RapidFuzz for slide number matching
• Gradio UI design

---

👨‍🎓 Author

Deshan Senanayake
BSc (Hons) Artificial Intelligence & Data Science
Robert Gordon University (via IIT, Sri Lanka)

---

🏁 Future Work

• Export flashcards as .csv or .txt
• Add MCQ quiz generator
• Add feedback loop to improve flashcards
• Deploy to HuggingFace Spaces or Streamlit Cloud
• Voice input (Whisper integration)

---

📜 License

MIT License – free to use, modify, and share with credit.