CollegeLoans

An all-in-one platform for College Planning.

Comment

Inspiration

Every year thousands of people have college debts far greater than they can pay. Our platform automatically calculates your debt and creates the best option for you with AI-powered services. The average US college debt is $35,000. The biggest cause of this debt is both financial capability and poor planning for the future. Our platform gets rid of this by suggesting students universities according to their SAT scores, location, and annual income.

What it does

Our goal for the user is for them to only focus on studying without unforeseen events, such as not finding a good college or going into a lot of debt.

CollegeLoans is a platform that lets you take a test. This test includes data like your region, SAT scores, and annual income. Based on this info, we create a profile for you that uses AI to recommend you your most compatible colleges and your debt. We use the annual income and your location to predict the debt, and we also use your SAT scores to recommend colleges to you.

How we built it

We used CollegeAI's API for the college recommendation part, which creates the recommendations based on the user's SAT scores and location. We also used Tensorflow to build custom regression algorithms to predict the user's debt taking into account a monthly interest rate. We used FastAPI to build an API that makes these models accessible for any web application, and we hosted the API on Google Cloud using their Compute Engine service. This made it way easier for us to just deploy our code and not worry about any usage spikes or the models being too heavy. We also use data from the College Scorecard dataset for the median debt per region based on the SAT scores.

The frontend was built with React and Tailwind CSS. The authentication was made with Auth0.

Data Usage and Modeling

As mentioned, we used 2 different datasets for our API. The College Scorecard dataset, which has data for the average debt per university and student on each percentile based on their SAT scores and their annual income. We connect this data to a Neural Network which predicts the debt based on the region of the university and its average debt. This data is then connected to the CollegeAI API which creates a dataset with all of the universities' basic info and cost. This API also powers our college recommendation section of the app. All of the universities are then displayed with our forecasted quotas and the predicted debt for the student based on their SAT scores, region, and annual income.

Challenges we ran into

Even if CollegeAI had a public API, there was no documentation for it, so we basically had to reverse-engineer their web application for our recommendation service to work. We also had some trouble creating the Virtual Machine and making it accessible as an API, but we, fortunately, made it all work out in the end.

Accomplishments that we're proud of

We're very proud of having a completed app with all of the features that we planned for it working correctly. We had experience with previous FastAPI instances crashing on the spot, which made us hardcode a key feature of our website. However, that didn't happen with this project, and we managed to get a professional full-stack web application with all of the features working. Hardcode is a common thing in Hackathons, but we managed to prevent that in this project, which might be one of the best our team has made.

What we learned

We learned a lot about API hosting on GCP and we also had never used CollegeAI's API before, so we also learned a lot about getting it to work correctly.

What's next for CollegeLoans

We're planning on adding support for all of the states in the US instead of being based on regions. We're also planning on adding more AI-powered services like adding support for GPA and ACT scores.

Built With

Share this project:

Updates