Mathplify

Inspiration/Problem

According to a 2019 survey by the National Association of Secondary School Principals, 63% of high school principals said their students felt unprepared for finals, and 57% said they had insufficient resources to provide enough practice problems. I also had a friend who was studying for finals and they didn't have enough practice problems and needed more. But the teacher did not have any more and it was difficult to find problems online, and we all know Practice Makes Perfect!

What it does

Students can upload an image of a math problem, both word problems and problems with equations, and the website uses AI to read the problem/extract LaTeX. It then sends the problem to another AI model that will tweak it and generate more practice problems.

How we built it

I used Python as a backend, Django as a web framework, Supabase to host a PostgreSQL database, and Vercel to host the website. For the AI models, I used MathPix's LaTeX extractor to read math problems and based the problem generator on the Bard LLM.

Challenges

It was difficult hosting the website on Vercel since Django is not usually used with Vercel, caused a lot of errors. I also had a lot of issues with the AI's accuracy in generating problems and formatting them properly so a LaTeX renderer could understand them.

What's next for MathPlify

Creating mobile apps powered by the same structure is probably the most important next step as it's much easier for students to snap a picture on their phone. Improving the model's accuracy is also important for very complex problems. Finally, after the product is polished, providing schools with this new tool can allow students to have unlimited practice.