cardmaster

Recommended Credit Cards
Wishlist
Learn More & Sign Up
Financial Advisor
promotional website: https://cardmaster-hackharvard.netlify.app/

Inspiration

Customer acquisition costs are high for credit cards. So high, that banks will even give out new iPads for anyone who signs up for their credit cards. On our flight from Montréal to Boston, we saw credit card ads plastered everywhere, offering sky-high sign-on bonuses. While banks are eager to reduce their customer acquisition costs, customers are bogged down by confusing terms and conditions, convoluted points programs, and choice fatigue.

With enough knowledge, customers can use the strategy of credit churning - obtaining multiple credit cards to capitalize on high sign-on bonuses and rewards. However, identifying the right credit card for oneself is usually a tedious process, and the card that a user selects might not fulfill all their specific needs.

What it does

CardMaster simplifies the credit card churning process, offering our users personalized credit card recommendations based on their demographics (occupation, income, credit score, spending habits and travel preferences), saving the time and effort required for extensive research. Apart from letting users reap the rewards, it also immerses them through mixed-reality experiences showcasing a view of the specific benefits cards’ offers would bring them. Put on your AR/VR headset and, with a few swipes, our "cardmasters" are virtually transported to the destinations you've been dreaming of visiting. They can virtually explore luxury resorts, see the world from your favorite airline's first-class cabin, and experience the perks of each recommended credit card firsthand.

How we built it

We built the mixed-reality front-end with AR/VR using the Apple XCode-Beta development kit, using Swift, SwiftUI, RealityKit, Alamofire, concurrent MVVM design architecture, and of course, VisionOS. We created a microservice backend, using Python, Flask, Bitarray, Scikit-learn, NumPy, Pandas, Google Cloud Platform (GCP), Cohere, and Docker. As a bonus, we built a demo website to showcase CardMaster, using JavaScript, React, and Tailwind at https://cardmaster-hackharvard.netlify.app/. Prior to developing the interface, we designed mockups using Figma, keeping simplicity, ease-of-use, and aesthetics in mind to create a seamless user experience. For added functionality, we also integrated an AI financial advisor which we prompt engineered for Cohere’s large language model.

The recommendation algorithm was accomplished using a Jaccard Similarity Index to find the similarities between users based on their demographics, such as credit range, annual budget, income, profession, and travel frequency. This was then mapped to the credit card tastes of each user. Linear regression was then applied to these results to determine whether or not each card was deemed a suitable fit.

Challenges we ran into

The VisionOS frontend environment was particularly challenging to work with due to the difference in the base components for VisionOS compared to conventional Swift components, which posed an interesting challenge for our UI/UX design. Additionally, navigating the VisionOS environment was also limited due to the lack of resources available online, as the Apple Vision Pro has not yet been released. We encountered difficulties in accessing support for many features we wish to incorporate.

One of the more difficult issues we faced was designing our recommendation algorithm, which could be thought of as a “black box”. We knew that persons with different credit score ranges, budget preferences, income, etc. would prefer different credit cards, but generalizing outputs to be statistically similar was at the core of our backend algorithm. Devising a solution to this involved ample research and randomized input dataset generation to train our model.

Accomplishments that we're proud of

Overall, we are very happy to have been able to work with the Apple Vision Pro, considering it has not yet been released. We’re proud of coming up with an innovative credit card recommendation algorithm and making the credit card rewards and application process more transparent.

What we learned

Our team had a diverse mix of experience from different backgrounds. The key takeaway from this experience was the ability to cross-collaborate on different ends of the project, teaching one another new technologies of the software development process. Some of us delved into the intricacies of Google Cloud databases, others explored unfamiliar frontend frameworks, and some of us dived deep into some cool math for our backend.

CardMaster turned out to be rewarding for all - our users would reap its benefits, while for us, designing and developing it was an immensely gratifying experience. We’re very happy with our completed project!

What's next for CardMaster

At HackHarvard, we used the opportunity to build CardMaster, particularly the backend services, to establish a foundation for a broader credit churning and optimization platform that we are developing. CardMaster churns credit cards, providing users for a call-to-action to improve their financial potential. The future phases of building CardMaster would address the management and optimized usage of a multi-card wallet for real-time transactions. Through these developments, our work will eventually improve the usage of credit cards at all stages of the churning process - before and after.