Bucket!

Fill your bucket with financial statements and watch as they turn into money you can control.

Comment

Inspiration

We were inspired by a certain team member's (that shall not be named) (Billy Campbell) current poor financial situation caused by their own bad decisions (allegedly).

What it does

Our project is composed of two parts, which both serve as ways to help students have more responsible finances. The first part is a chatbot that helps visualize a user's finances into "buckets", asking probing questions and giving example budgets to help the user find the best budget for their situation. The second part is a tool to read a pdf bank statement and sort it into categories, allowing for an LLM to make comments and suggestions on how a user could better manage their finances.

How we built it

Our webapp runs on Python with Django as the backend. Bucket! uses custom bucket data structures to represent the user's budget in the backend, with Chart.js to help visualize the budget. The chatbot runs on OpenRouter with Google's Gemini 2.5 Flash LLM and custom prompts to better suit our usecase. Vellum uses pdfplumber to read the uploaded PDFs and send their output to Google's Gemini API. Our code was written on the Cursor text editor.

Challenges we ran into

We initially struggled with choosing a direction for our project, which is what had us divide into subteams to develop both the chatbot and pdf tool. When working with the LLMs, we had to deep dive into the best prompting techniques to allow us to get our desired results, and we spent hours troubleshooting our prompts to make sure each part of our process can provide useful and usable output.

Accomplishments that we're proud of

We're most proud of how we fine-tuned our prompts to work well with one another, as we came into the hackathon with our main goal being to get up-to-date with LLMs and LLM-integrated development

What we learned

We learned a lot about working in a team on a project, especially when it comes to working with git in a multi-user environment. We learned a lot about new AI technology from trying Cursor for the first time, and attending multiple related workshops.

What's next for Bucket!

We plan on expanding out the chatbot side, integrating a database and RESTful API for potentially integrating the budgeting side into a future project.

Built With

Share this project:

Updates