Inspiration
After working at an accounting office and working hours on end bookkeeping, Blake, one of our members, wished there was a more efficient way to summarize and read bank statements. We all agree that digesting bank statements can be tough, so we wanted to see if we could use an AI model to help categorize and neatly present our spendings.
What it does
Motion Finance is a website that takes in bank statements and presents the user with a comprehensive look at individual categories of spending. Once a bank statement is submitted, spending categories are automatically applied to each transaction. The user can also ask an AI chatbot for tips on reducing spending.
How we built it
Starting the project, we drew out a concept for a website and visualized how we wanted to present our data. We built a Flask app using HTML and CSS to create a responsive front-end, complete with a sign-in page and multi-route functionality. In Python, we wrote a script to parse bank statements, filtering out unwanted data, and then storing it in MongoDB Atlas as our backend to be categorized by a custom fine-tuned AI model. Our website receives this processed data, presenting it into static scrollable tables and a pie chart. Lastly, we integrated our AI chatbot into our website that responds to live changes in database content.
Challenges we ran into
We ran into a few roadblocks while creating this project. We started building the app on Django, but a hefty effort to facilitate MongoDB connectivity lead us to opt for Flask instead, which offered native MongoDB support. Connecting all members on to the MongoDB database proved to be difficult. We did not realize that the eduroam network blocked our access, taking a hit on our productivity.
Accomplishments that We're Proud Of
After some of us worked with a local SQL database before, we wanted to take it to the next level and attempt online database connectivity! We're proud that we managed to work out our connection issues and actively modify and query the Atlas database. Additionally, learning how to fine-tune an AI model sounded super scary, but a careful read of documentation and a bit of patience came together for a pleasing outcome to utilize in our app.
What We Learned
On top of our ventures with Atlas and model training, this project allowed us to develop a more thorough understanding of Flask application development, including efficient file structure, and how to distribute workload when collaborating with others. We all showed the best of what we could do and counted on each other to fill the gaps where we'd fall short, a dynamic that goes a long way in creating a product everyone is satisfied with.
What's Next for Motion Finance?
A lot! We're going to continue polishing the front-end to make it look as presentable as possible while we workshop new ideas and concepts into Motion to improve its usefulness and functionality. Similarly, we want to implement a live service that can allow for users to talk and see plans for their budgeting.
Log in or sign up for Devpost to join the conversation.