Branchless

• 

  Dashboard

• 

  QR Code

Inspiration

We were inspired by the impact of the lack of financial services in many rural locations, which could lead to major negative ramifications on the lives of many. A digital bank would solve many of these issues, and enabling user-to-user transactions with less hassle would be especially beneficial.

What it does

Branchless serves as a digital wallet with several distinguishing features. Firstly, it can deposit checks virtually through an LLM that parses in an image of the check. Through this, it's able to efficiently figure out Next, it also has highly accessible user-to-user financial transactions, which can be conducted with just a QR Code or Account Id. It uses 2-Factor Authentication (through email) for increased security as well.

How we built it

We built it in next.js, using tailwind css for styling. We used the llama 3.2 Large Language Model to handle the check deposition parsing. We used Supabase as our database to store information, as it also helped facilitate 2 Factor Authentication.

Challenges we ran into

At first, we struggled to limit the runtime of the LLM to a reasonable length, with it's initial iteration requiring over 2 minutes to succeed with accuracy. However, through preprocessing the LLM initialization, we were able to cut down on the runtime by a massive factor, reducing it to less than 5 seconds on average.

Accomplishments that we're proud of

We're proud of our check parsing accuracy, as it's maintained practically 100% accuracy despite it's relatively quick speed. We spent a lot of time and effort on this, so it was rewarding to see it pay off.

What we learned

We learned that when under a low time constraint, we should figure out alternate strategies to speed up. In our case, this was altering how we split up working, finishing one feature at a time instead of individually working on different things.

What's next for Branchless

We plan on improving integration for actual monetary deposits. Tools such as Stripe would be the most effective way to implement this. We also plan on implementing a monetary extraction feature,

Built With

• javascript
• llama
• next
• node.js
• postgressql
• react
• superbase
• tailwind