EdAI | Devpost

Inspiration

I ran into a rather pressing issue at school. I noticed that students struggle to stay on top of their assignments, homework, and group projects, while juggling extracurriculars. In this competitive environment, students are in a constant rat race, barely scraping by. They don't know how to deal with the amount of work being given to them, and instead need to cram and lose sleep to get everything done. Often, these students lack the resources needed to accomplish certain tasks in a timely manner, and resort to softwares like ChatGPT to solve problems, write portions of their essays, and more.

What it does

This is where EdAI comes in. EdAI will be designed to help students learn, and instead of just giving them answers, which does not benefit them at all in the long run. Our pre-trained AI model acts almost like a teacher. EdAI is programmed to lead students to the answer based on their own previous knowledge, and instead of just giving the answers. The app reaffirms knowledge that the student already has, and nudges them towards the correct answer. EdAI is also able to boost a student’s mental health by using AI to tell the user a joke, and uses a generative AI image model to fabricate an image that goes along with the joke.

How the app was built

We built the app using the Flutter framework. We used OpenAI API to make generative text models that would return a response to the user’s query. We also used the API to use the Dall-E2 image generation.

Challenges I ran into

One of the biggest challenge that I ran ran into was learning the specifics of Flutter. I knew the basic framework and syntax, but I had to quickly learn how to connec the dots. There were also lots of bugs that showed up throughout the development process, which I took lots of time to resolve.

A second challenge occurred during the testing phase of the app. We weren’t getting the API request to go through, so I had to try to find out if it was an error on our side, which it was. Then, I had to go through the documentation again to find the root of the issue. It turns out that the versions in the API configuration file did not match up, so I needed to make sure that we were using the correct models.

Accomplishments that I am proud of

First off, the idea of changing how AI interacts with students was pretty neat, as students now can simply pull up a tab of ChatGPT and get the answers. We are able to change this by empowering students by showing them the step-by-step process to crack those tough problems. It's like having a personal tutor who guides you through the maze of assignments, ensuring you not only find the solution but also understand the logic behind it.

Another accomplishment that I am proud of was training the model that is being used in the app. I was surprised that the model followed the instructions verbatim, but was still able to guide me to the answer. I decided to feed it some problems from my homework as training data, and it ended up handling the problems seamlessly. However, The AI did have some premisconceptions about certain math topics, but I was able to fix this easily.

What we learned

I learned a lot about the Flutter framework in general. I would describe myself as a beginner prior to the project, but by putting myself through a project with lots of adversity, I was able to push myself out of my comfort zone and learn infinitely more. I was able to use version control to solve the various bugs that I came across. Cominbing this by using open source resources, I was able to grow as a debugger this way I also learned how API calls can be utilized throughout the program, and what errors I might come across.

What's next for EdAI

For our app, we plan to give teachers the liberty to make calculated decisions using the app. We plan to make a backend where the teachers are able to enter their credentials to login. Once inside, teachers can navigate through a user-friendly interface that puts them in control. They can see what common questions students have asked, which assignments students have used EdAI the most, and the chat history of a certain user. That way, teachers can create curated lesson plans so that students have their doubts about particular assignments resolved. In addition to this, during a certain time an assessment is administered, the teacher will have the ability to disable EdAI so that students aren’t able to access the app to mitigate any instances of cheating.

Built with (What languages, frameworks, platforms, cloud services, databases, APIs, or other technologies did you use?):

Dart, flutter, openai, chatpgt, dalle