CredRec | Devpost

Inspiration

As financially struggling college students who felt overwhelmed with the vast depth of knowledge and study required to properly manage our finances, we felt there needed to be a tool to help people make smart financial choices. The particular choice of selecting your arsenal of credit cards amongst the seemingly endless options can now be made simple!

What it does

CredRec analyzes your financial data usage and scrapes the web for credit card benefits and gives you recommendations on the best credit cards for you. By querying the LLM, users will be able to ask questions about credit cards to find a credit card that would be the best fit for their current needs and worries.

How we built it

CredRec connects user data involving transaction history with Google's Gemini LLM API trained on the web-scraped data of credit card offers.

Challenges we ran into

We faced difficulties during web scraping due to the website we chose having a mix of dynamic and static structuring as well as obstacles that prevented methods we employed to scrape data, forcing us to develop new code for the challenges we came across. We also initially wanted to use real transaction history from users with Plaid API but discovered that access would only be granted through an approval process that we did not have time to complete, opting to use Faker.js instead to mock data for our purposes.

Accomplishments that we're proud of

We are proud of the project concept we came up with, as we strongly believe in its real-world applications when refined. We overcame a plethora of unexpected issues throughout the project (learning hundreds of methods that didn't work in the process of finding solutions). Everybody managed to learn something new in the process, and persevering through 24+ hours of coding without sleep was the biggest thing we could be proud of! We also became really good as a team at quickly switching between tasks in order to assist each other and guaranteeing a fresh set of eyes when hitting roadblocks and ensuring everyone gets some understanding of every aspect of our codebase.

What we learned

We learned the process of setting up Git environments in a team setting, working through frontend and backend development, developing a web scraper on a website with suboptimal formatting, and the utilization of many different software and APIs to conjoin in the development of the project. We learned that when building projects on a timeframe, it is best to use a more focused tech-stack. Early on in development we were casting too wide of a net trying to do too much which resulted in constant design reworkings. With our groups varying skill levels, we learned both how to learn and teach efficiently while dividing and collaborating on tasks. Spending time explaining both our own code and the problems we were facing to each other was extremely powerful, as it both solidifies our own understanding and opens the path to creative unexpected solutions.

What's next for CredRec

CredRec would develop further by utilizing web-scraping and cloud services to provide the most up-to-date information about credit card deals and limited-time offers. With this, we could offer users with even more curated information and recommendations for their needs. We would also develop the front-end further to include user history on their queries and implement the Plaid API to replace Faker.js with real data for user-specific transaction history to provide higher-quality recommendations.