Intelli Learn

An AI homework helping site that accommodates visual learners to help them get better at math.

Comment

Inspiration

Many mainstream AIs simply provide you with an answer to a question you provide it. Sometimes, their response may be correct, while others, not so much, and without much experience/confidence in the type of question(s) you provide it, you may never know if it's correct. Many language models are good at shaping incorrect answers to appear correct, which is problematic. Our approach is much simpler: help younger children with their math problems. But this isn't just a "type a question in the textbox and get an answer". This is more than that. Many parents who help their elementary school-level children with their homework have to do the tedious task of drawing out the problem for them to better visualize it. Some children are just born visual learners and can't grasp the question without a picture presented. This AI does just that. Instead of text that walks you through it, it provides easy-to-digest images of a simple word problem you provide it to save the hassle of parents having to draw the problem out and waste paper; they can simply type the problem in and have the image generated in an instant!

What it does

It helps to solve high school math word problems in general (it can do lots of other things), by generating explainable images, which is the selling point feature of our AI homework helper. We found it better than even ChatGpt 4 image generation tool. It can also help with basic English grammar questions like meanings, synonyms, etc.

How we built it

We used React and TypeScript with APIs and datasets of questions (123) for quizzes (adaptive and randomized every time) and AI help for homework. We used Git as a collaboration tool (took 30% of the time). Heroku served as a proxy for bypassing CORS policy for our API. We first created our UI (using React + Vite). And I was working on the API and other backend stuff alongside solving conflicts in repos. And Deployment on vercel.

Challenges we ran into

For the most part, we struggled with Git and Github. It was a tedious challenge in the beginning, but as we worked through it, occasionally ran into the same issues, and thanks to ChatGPT, it made things smoother. Of course, there were periods where we struggled quite a lot with big Git conflicts. Another problem was the AI API for the word problem to image generator. It was really hard controlling the input and preventing this white error screen from appearing when there were conflicts as we intended to let the user know there was an error rather than display this "white screen of death". We faced issues with score manipulation and storing and passing to other components due to state. I solved that issue. Then we did try to use game theory for quizzes. The API endpoint was working with the browser but failed with our application, giving us a CORS policy error. I came up with a solution to solve this around which I learned recently using Heroku as a proxy gave us a way to come around that problem. Then API response tailoring according to our requirements took time and effort. Overall, we were able to incorporate maximum features we can in this short time frame. Last minute deployment was the best challenge we had.

Accomplishments that we're proud of

We made a workable solution that can be used by people; we just need to deploy it soon with a few changes for security. Integrating APIs and creating an adaptive randomized quiz. We got better at understanding Git and GitHub.We were able to deploy our site at last minute on https://hen-hacks2024.vercel.app/

What we learned

We learned to collaborate, learned React and TypeScript which were new to us, but having familiarity with JavaScript helped to take it on as we go. Learned to Use OpenAI tools and other APIs which we used and some we did not use. Got to know lots of education tech ideas during our brainstorming process that opened us to lots of ML applications in the field of Education. We then checked the feasibility of the ideas with our skill set and also moved way beyond our capacity and took the challenge and it worked out well and rewarding. It was a first live deploy of web application .The hackathon was really enjoyable and full of experience. Made lots of connections, took lots of help. Thank you Team and UD (all other people who helped us directly or indirectly) for this wonderful learning experience.

What's next for Intelli Learn

We want to introduce more subjects like Science, Social Studies, Languages, etc. Taking voice input and giving out voice output which can increase more accessibility. We could not do that due to time constraints. Making it better by training it more with good datasets. Gamification of the process of learning can be the best next step that will further our project. User behavioral analytics can also improve project quality.

Built With

Share this project:

Updates