Movin

Inspiration

Choosing college housing involves more than comparing rent. Utilities, deposits, moving costs, roommates, and changes in income can all affect whether a lease works for a student.

We built Movin to answer: “How would this housing choice affect my finances throughout the school year?”

What it does

Movin is an AI financial copilot that combines fictional banking data from Capital One’s Nessie sandbox API with campus cost estimates.

Students can describe an apartment through chat, including rent, roommates, lease dates, and move-in fees. Movin calculates monthly housing costs and upfront expenses, forecasts their balance throughout the lease, and flags when they could fall below a chosen safety buffer.

The interface also offers neighborhood comparisons and multiple school options. School profiles and listings are part of the hackathon demo, rather than live apartment inventory.

How we built it

We designed the interface in Figma and built the web application with Next.js, React, TypeScript, and Tailwind CSS, deploying it on Vercel.

Our TypeScript financial engine combines banking inputs, campus estimates, housing costs, and scheduled cash flows to calculate forecasts. We added a machine learning model that learns weekday spending patterns when enough history is available. It is selected only when historical testing shows it performs better than a simple average; shorter histories use the average-based forecast.

Fetch.ai’s ASI API helps interpret conversational housing details. The financial engine calculates the results rather than relying on the AI model to invent figures. We also built a Python agent integration with Agentverse and ASI:One.

We collaborated through GitHub and used Codex to assist development and testing.

Challenges we ran into

One challenge was connecting the frontend, banking integration, campus data, and financial engine while working on separate branches under a deadline.

Another was making conversational inputs reliable. The agent needed to distinguish total apartment rent from a student’s share, preserve details across messages, and ask for missing costs rather than assume they were zero.

We also needed to communicate the forecast’s limitations clearly, especially when spending history was short or future income had not been scheduled.

Accomplishments that we’re proud of

We brought our separate components together into a deployed application that turns housing details into financial estimates.

We’re proud of the conversational experience, the cost breakdowns, and the forecasting approach that compares a learned spending model against a simpler baseline. We also demonstrated our agent directly through ASI:One.

What we learned

We learned how to connect APIs, build and test financial calculations, integrate an AI agent, and collaborate across different parts of an application.

We also learned that useful financial predictions depend on clear assumptions and reliable inputs—not just adding an AI model.

What’s next for Movin

We want to improve and verify campus data for more schools, support richer income schedules, and evaluate our forecasts using longer spending histories.

Our goal is to keep developing Movin into a tool that helps more college students understand the financial impact of their housing choices.