Inspiration

Creating Dutchie, we aimed at solving a problem that we encountered surprisingly often - having to worry about the burden of splitting the expenses after or during an exciting group event. Counting spendings and paying each other is really not what you want to think about after a boat trip or a bachelor's party.

What it does

Dutchie works with whatever context you can feed it - transactions records, bills and receipts, photos of the dinner table, group selfies, or even just a voice/text message descriptions. Processing all that data, it recognises how many people participated in the event and their expenses. Based on that it builds a visual representation (a graph) that clearly shows who payed for what and who owes whom. Users can then manually adjust the scheme according to their considerations, and then Dutchie sends payment requests automatically upon all users' confirmation.

How we built it

We built Dutchie as a FastAPI web app with a small browser UI as an MVP of the project !

Challenges we ran into

The hardest part was turning messy data like photos, receipts, and messages into clear expense rules and dealing with various operation scenarios. We also had to keep AI interpretation separate from the final money calculation.

Accomplishments that we're proud of

We built a working pipeline from raw event context to a visual expense graph, and what we personally like the most, is that it is super intuitive and "at your fingertips", saving users lots of time and hustle.

What we learned

Leveraging multimodal AI showed us that social debt doesn't have to be complicated. It is indeed surprising how effectively we could automate the transition from messy and unpleasant group dynamics into a structured, understandable payment scheme that just works.

What's next for Dutchie

Within the bunq hackathon 7.0 timeline, we have developed a MVP of the project, showcasing the core idea and key principles. Further on it of course may be scaled and developed into a production-ready, cloud-native integrated platform.

Built With

Share this project:

Updates