KRAM

Using AI to generate practice problem sets for any subject

Comment

Inspiration

We wanted to create a study tool to help students go beyond their study guides and homework. When studying concepts, it can be difficult to find example problems to practice with.

What it does

KRAM allows users to upload images of their homework and worksheets to generate practice questions covering similar concepts.

How we built it

We used React Native and Tailwind CSS for the frontend. Firebase was used for authentication and storing user information. Google Vision scans images for text. Google Gemini creates the multiple choice practice questions.

Challenges we ran into

Google Gemini is not the best with computation, so we ran into some issues with math practice problems. We pivoted more towards more conceptual practice problems, but we have some ideas to implement some form of computation checking to get correct answers from Gemini's recommended problems.

Accomplishments that we're proud of

This is our first hackathon, so we are proud of how much progress we were able to make in just a few days. We had a slow start due to lack of experience with some of the technologies, but we were able to get into a good rhythm and come up with a solid product.

What we learned

We learned a lot about mobile development (React Native, Firebase, TypeScript) which will be very useful in future projects. We also learned more about prompting LLMs and how to better obtain responses that fit our specific needs.

What's next for KRAM

We plan on continuing development and preparing KRAM for release on the iOS app store and possibly other platforms.

In future development we will focus on optimizing the LLM prompting to ensure accuracy and relevancy in the questions and answers it returns.

We have some ideas to help gamify KRAM and keep users encouraged while studying. We also would like to implement various settings to help users customize their studying experience. For example, increasingly difficult problems as user accuracy improves.

Built With

Share this project:

Updates