• 

  Base code for development of EduMate an AI-Powered Learning Assistant

• 

  Frontend Deployment

• 

  Frontend Deployment

• 

  Code for UI

• 

  Code for Vector Store to processing and storing of data

• 

  Code for testing AI Responses

🚀 AI-Powered Learning Assistant – Project Story

🔍 About the Project In today’s fast-paced education landscape, students often struggle to find timely help outside the classroom. Inspired by this gap, we built EduMate — an AI-powered learning assistant that acts as a personal tutor, capable of:

1. Answering subject-specific questions, 2.Explaining complex concepts with examples, 3.Summarizing lecture notes and textbooks, 4.Adapting responses based on a student's learning level. Whether it's midnight math doubts or a quick review before exams, EduMate is your go-to study partner.

💡 Inspiration

The idea came from our personal struggles during exam season — juggling multiple subjects, unstructured notes, and limited time for revision. We imagined:

“What if a smart tutor could read our notes and instantly help us understand what we’re stuck on?” This thought led us to leverage LLMs (like GPT-4), OCR, and context-aware summarization to create an assistant that actually learns from your materials.

🛠️ How We Built It

📚 Tech Stack

1. Frontend: Streamlit (for fast interactive UI) 2.Backend: Python 3.AI Models: OpenAI GPT-4 via API, HuggingFace Transformers 4.OCR: Tesseract for scanned handwritten notes 5.RAG: Retrieval-Augmented Generation for context-based answers 6.Storage: Pinecone / FAISS for vector database (contextual search) 7.Document Processing: PyMuPDF and LangChain for note ingestion

🧠 Core Features

1.Question Answering from User Notes We parse uploaded PDFs and build a vector database for fast similarity search. Queries are embedded and used to retrieve relevant note chunks. The LLM then answers based on both the query and the context. 2.Topic Summarization Summarize long notes or textbook chapters using abstractive summarization. Output simplified explanations in bullet points or step-by-step format. 3.Concept Explainer (with Math support) Explains formulas like ( E = mc^2 ) or derivations using LaTeX formatting. Converts images of equations into readable explanations via OCR + AI.

🧗 Challenges We Faced

1.OCR Accuracy: Handwritten notes were noisy and error-prone. We mitigated this using preprocessing (binarization, denoising) and fallback to typed text input. 2.Chunking Documents: Chunking long PDFs in a semantically meaningful way was harder than expected. LangChain helped, but we had to tweak chunk sizes and overlap for quality answers. 3.Token Limits: LLM token limits forced us to intelligently narrow down relevant content before querying the model. 4.Context Switching: Maintaining topic coherence across multi-question conversations was tricky without session memory.

📚 What We Learned

1.How to build a RAG-based AI system that feels contextual and personalized. 2.Optimizing prompts to improve educational value and clarity. 3.Using OCR and PDF parsing in real-world educational tools. 4.The importance of UX in making AI actually useful for learners.

🌟 Conclusion

EduMate isn’t just a chatbot — it’s a personalized AI tutor that evolves with your learning. Whether you're studying calculus or reviewing history, EduMate helps break down knowledge into bite-sized, understandable chunks — anytime, anywhere.

Built With

• matplotlib
• numpy
• ocr
• openai
• opencv
• pandas
• python
• streamlit
• tesseract