Lumina

Inspiration

Students spend years studying yet often graduate without true mastery of the material. Professors upload dense PDFs and long slide decks that students read once and quickly forget. We imagined a world where every module in every course could transform itself into a personalized learning environment built around clarity, retention and real job-ready skills. Lumina was created to make that world real by turning traditional course content into an intelligent and adaptive academic companion.

What It Does

Lumina redefines course engagement through a seamless three-layer ecosystem powered by AI. The first layer is the Academic Foundation. Students experience a modern and intuitive interface with clear syllabus navigation and real-time tracking of their progress across both modules and entire courses. The second layer introduces AI generated study tools that are created automatically for every module a professor uploads. Lumina produces mind maps that reveal conceptual connections, flashcards that convert readings into quick review material, adaptive quizzes that adjust to a student’s understanding and a grounded AI chat system that answers questions using only instructor supplied materials with transparent citations.

The third layer enhances learning through interactive features. Students can listen to readings through integrated text to speech, view optional AI generated explainer videos and stay motivated through a gamified system of streaks, experience points and mastery rewards. Every tool is trained only on professor provided content to maintain academic accuracy and integrity.

How We Built It

The frontend is built with React, Tailwind and Lucide to achieve a clean, responsive and organized learning interface.

The AI and backend architecture, currently in development, uses Gemini 1.5 and Gemini 2.5 models along with the Gemini Files API to process PDFs, slides and videos. PGVector powers the embedding system, Vertex AI Search handles retrieval, Cloud Text to Speech enables audio learning and PostgreSQL stores student progress and course structure. Instructor uploads move through a pipeline that extracts, embeds and organizes content into module specific knowledge that fuels each AI tool.

Challenges

One major challenge was designing an AI system that remains strictly focused on individual modules rather than producing generic answers. Integrating Gemini’s multimodal capabilities within a tight timeframe required careful engineering. Maintaining academic integrity demanded strong separation between study materials and graded assessments. Another challenge involved creating smooth navigation across the full hierarchy of dashboard, courses, modules, learning tools and document viewers.

Accomplishments

We built a functional prototype containing four fully operational AI study tools within a cohesive interface. The platform supports live tracking of module and course progress, grounded AI chat with reliable citations and integrated text to speech directly inside document viewers. The architecture is scalable and ready to be expanded into a full university level system.

What We Learned

Module specific context improves AI accuracy dramatically. Students place greater trust in AI when answers include citations. Gamification increases motivation and long term engagement. We also learned the importance of prioritizing essential features when building under strict time constraints.

What’s Next for Lumina

Our next steps include forming partnerships with universities that want to modernize their learning systems. We plan to add AI generated videos and micro podcasts, build comprehensive instructor analytics dashboards, enhance accessibility features, create a large scale database of courses and instructors and expand Lumina into a full mobile application.

Tech Feedback

Gemini’s multimodal capabilities were extremely powerful and intuitive. The MLH React and HTML workshop was challenging but provided valuable foundations for structuring our user interface.