Eduverse

Inspiration

A lot of us have experienced understanding a topic and finding the lesson easy and getting blindsided by the difficulty of the next topic. Eduverse was created with the goal of adaptively adjusting content type, question difficulty and pace.

We believe that learning and moving on from a topic should not be defined by simply an assessment score but rather should holistically and effectively assess the student's understanding of a topic

What it does

Eduverse is a learning management system that keeps teachers in control of designing lesson plans and identifying links between topics. Teachers also control exactly what questions are shown to their students when creating assignments. Lastly, teachers also upload multimodal learning materials to aid students in actively revising while attempting questions.

Students are presented questions of moderate difficulty at first and based on their performance, the difficulty of the questions are dynamically adapted.

How we built it

Eduverse is built using NextJS - a full-stack React framework that allows for rapid bootstrapping of code and ease of setting up endpoints - alongside TailwindCSS.

The user data, authentication, questions, classes and other data are stored in Supabase.

The AI that powers this application is known as Deep Knowledge Tracing - a Recurrent Neural Network (LSTM). This works by modelling a student's understanding of a question by observing their interactions with questions and then using the information to predict their performance on the next question.

This AI model was implemented using PyTorch and used FastAPI to expose the endpoint to the main application.

Challenges we ran into

We ran into challenges while creating the linking structure of notes and creating a intuitive interface that would allow teachers to seamlessly plan courses. We also faced issues persisting this information to the database such that it could be retrieved and used in a meaningful manner.

Secondly, we faced issues training and testing the Deep Knowledge Tracing (DKT) model due to limited datasets available.

Accomplishments that we're proud of

We are proud to have a fully functional app with a relatively clean user interface. We are especially proud of the Lesson Flow feature which allows teachers to indicate dependencies and foundations of topics - allowing our application to dynamically adjust the content type.

What we learned

We learnt how to integrate different tech stacks and apply our knowledge across disciplines to create this app.

What's next for Eduverse

We wish to enhance the existing gamification in Eduverse

Built With

• fastapi
• nextjs
• pytorch
• tailwind