CATLOLL | Devpost

Inspiration

The main inspiration for creating a mortgage advisor from the spinning cat meme was being able to bring my two favorite worlds, cats and credit assessments into an application that is hellbent on changing the world. I knew the moment I locked eyes with the cat, I knew it's future would be best suited in finance. It's not everyday you see a cat give out a 30-year fixed-rate mortgage without clawing you with non-flexable loan options. I once joked with a friend about how even a cat could do a better job explaining mortgage options than some advisors. This is why I don't have that many friends. Ultimately, I wanted to make something as boring sounding and complicated as home financing into a more accessible medium. The spinning cat gave me the proper flavor to create a fresh interface for a more mundane topic.

What it does

Sometimes it's what about what it doesn't do. This cat will not give you any conflicting loan terms due to the fact it's not incentivized to think of only the cat's profits. It prioritizes great transparency and trustworthiness, offering loan options that align with the user's responses and preferences. The cat takes user inputs, such as income, credit score, and housing preferences, to generate the best tailored loan recommendations based on your data. The interactive elements with playful animations or engaging questions make the process both enjoyable and informative for the end user. The application also effectively educates the users about mortgage terms, interest rates, or loan structures in an easy-to-understand way. The system ensures transparency by showing the most detailed breakdowns of interest rates, monthly payments, or hidden fees .

How we built it

We built this app by using Next.JS and Python to create the most intuitive user experience possible with the cat. We chose Next.js for the frontend because both of us knew React and also the server routes made working with LLM APIs and our Python backend very simple. Other services we used was Codebuff for generating most of the UI and Nebius for LLMs. We then deployed the application on Vercel where we pointed the domain catt.lol towards it.

Challenges we ran into

Challenges we ran into was trying to make the cat dialog system work correctly as well as fine tune the system prompts to give us the correct questions. One notable challenge was working with the video tag in HTML5 to display the cat. We removed most of the jank when working with it but it's still barely working on different type of devices like mobile. Most issues my team ran into like the minor api issues and UI bugs were always fixed relatively quickly and we worked tirelessly to get a working version of this application out to our proper vision.

Accomplishments that we're proud of

We're proud of how the vision and dialog system turned out as well as how dynamic and responsive the UI is working with LLMs and user responses. In particular the smoothness the animations and certain browser effects is something this team is greatly proud of. Just having all of the moving parts work together to the point of deploying the app is more than satisfactory.

What we learned

We learned how to properly deal with vision models and how to create a proper LLM conversational dialog system with multiple choice answers. We learned about many wonderful services and developer tools like CodeBuff and Nebius, and we plan to use them heavily in the future.

What's next for CATLOLL

Improvements to CATTLOL would likely come from more authentication potential to show dashboards, old conversations, updated interest rates, and more. Theoretically we could fine tune more specialized machine learning models for greater accuracy of mortgage advising. We could broker deals to compete with mortgage service juggernauts like Rocket Mortgage and SoFi. We're more than open to explore collaborations with banks or credit unions to offer real-time mortgage rates and pre-approval processes directly within the app. To keep up the fun nature of the application we could possibly add achievements for improving credit scores or interactive budgeting tools, to keep users engaged. I own this domain for a year so I might as well over engineer this cat even more somehow.