LUMI

Inspiration:

Education is the foundation of growth, yet millions of students still struggle with complex, inaccessible content. As a data scientist and artist passionate about AI for impact, I wanted to create something that makes learning easier for everyone — from simplifying heavy academic texts to automatically generating quiz questions. That’s how LUMI was born — short for “lumen”, meaning light. LUMI’s mission is to illuminate understanding.

What it does:

LUMI is an AI-powered educational assistant that: Simplifies complex study materials into clear, easy-to-grasp language. Generates adaptive multiple-choice questions (MCQs) to test comprehension. Offers a lightweight, privacy-friendly interface built with Streamlit. Runs fully in Python and works locally or via Cloudflare, making it accessible anywhere.

In short: LUMI acts as a mini tutor that rewrites lessons and creates instant quizzes to reinforce learning.

How we built it:

Language: Python Framework: Streamlit for the interactive web app NLP logic: Regex-based text simplification rules (extendable with Hugging Face transformers) Deployment: Cloudflare Tunnel for public access Development environment: Jupyter Notebook & Google Colab

All functionality is packed into a single, optimized notebook (LUMI_single_notebook.ipynb) for quick testing and submission.

Challenges we ran into:

Streamlit’s server couldn’t connect directly in Colab — had to configure Cloudflare tunnels. Debugging imports between notebook and app versions. Designing a clean UI that loads fast on limited resources. Ensuring reliable output even without heavy AI models.

Each hurdle helped refine LUMI into a more stable, elegant solution.

Accomplishments that we're proud of:

Built an end-to-end working prototype entirely in Python. Deployed a live Streamlit app from Google Colab using Cloudflare. Designed a beautiful 3D neon brand identity. Created a tool that actually teaches and engages learners — not just analyzes data.

What we learned:

How to bridge NLP logic with simple UI tools like Streamlit. The power of lightweight, privacy-friendly AI. How critical good UX is for educational technology. Cloud deployment and tunnel configuration in constrained environments.

What's next for LUMI:

Integrate transformer-based summarization (BART / T5) for deeper understanding. Add multi-language support (English, French, Arabic). Enable voice interaction and quiz export features. Create a mobile app version for offline learning. Build an adaptive learning system that tracks progress and personalizes lessons.

Built With

• fastapi
• github
• googl-colab
• jupyter
• natural-language-processing
• numpy
• pandas
• python
• random
• regex
• streamlit