Lumen

Inspiration

Lumen was inspired by our own struggles with studying and the inconveniences of searching for knowledge across disparate platforms. As students overburdened with coursework, pre-class materials, syllabi, recordings, practice tests, etc., it's nearly impossible to manage a limited screen cluttered with dozens of tabs just to find one useful resource to solve one question. Not to mention, this is not a one time problem; students must inefficiently repeat the search process for every question. What's more is that when students use traditional AI systems, that lack the essential understanding of the top-down and bottom-up connections between topics. Usually, they can only solve individual problems, and it is difficult for them to grasp the connections to other knowledge points.

What it does

To save users time, increase convenience, and help students better understand what they are studying, we developed the Lumen web application that runs as a study assistant and planner. Users can make a query for solutions they want to get from the course content library (Canvas, Panopto, MyPlan, and Gradescope). Lumen can generate study guides from course materials, offer step-by-step guidance through difficult problems, summarize lectures, readings, and notes for review, and keep track of assignments and deadlines across platforms. For example, a query like "explain to me what an implication is" will track sources from the CSE 311 syllabus and ED content. Then, the user will receive a reply about how implications work for this class and several supplemental explanations to help them better understand. Lumen can also make a study schedule for a user based on previous problems asked by them to help them consolidate their understanding of their knowledge weaknesses. After the system outputs the result, the AI can guess what the user wants to do next and give suggestions, such as making a review plan based on questions asked, providing a mock quiz, etc.

How we built it

We started by designing Lumen in Figma and animations in Lottie. We built the project using JavaScript, React, Vite for the front-end, and Python, PyTorch, Node.js, RAG, and AWS for the back-end.

Challenges we ran into

We ran into a lot of problem. A lot. The main challenge was During the project, the challenges we ran into centered around balancing development speed across our various responsibilities and properly managing AWS IAM roles and permissions across multiple services. We also faced the challenge of designing responsive UI animations under time constraints and had to work to maintain stable synchronization between data ingestion and retrieval.

Accomplishments that we're proud of

We successfully built and deployed a RAG pipeline on AWS that ingests and queries data from multiple university platforms, including Canvas, Panopto, and Gradescope. This system allows our assistant to generate personalized study guides, summarize various course materials, and create mock quizzes based on a user's identified knowledge gaps. We developed a complete, end-to-end web application, starting from the design in Figma to implementing a responsive React front-end and a functional back-end, while also solving data synchronization challenges to ensure a reliable user experience.

What We Learned

Throughout this project, we learned how to implement a RAG pipeline on AWS tailored for domain-specific contexts, which clearly demonstrated the power of contextual retrieval in improving AI reliability. Beyond the technical implementation, this experience also reinforced our understanding of the value of human-centered design used to solve true problems and its positive affect in society.

What’s Next for Lumen

For the future, our primary goal is to expand Lumen across multiple campuses and universities. We also plan to introduce new features, such as calendar synchronization for events, assignments, and study sessions, along with personalized notifications and reminders before tests, and many more features designed to help students stay organized and efficient.