Inspiration
Our inspiration came from our personal experiences. As we grow out of childhood, we have to start worrying about our own financial well-being. Thus, we thought it would be helpful to have someone who monitors our spending and provides real-time feedback.
What it does
The core feature of InsigntWallet is the ability to process bank statements into individual spending and provide ratings and suggestions on how each of the spending categories could have been done to obtain more value.
How we built it
We built a webapp using React frontend, FastAPI backend, and Cohere to build this project. Furthermore, we fine-tuned the CoHere model using financial datasets on Hugging Face to tailor the answers to suit our questions.
Challenges we ran into
Some of the challenges encountered include difficulties obtaining clean and usable data. This occurred both when we tried reading information from a bank statement and obtaining data to fine-tune our LLM. To clean the data, we leveraged pandas and numpy to remove unwanted information and labeled our data to make it usable. Another challenge that we encountered was issues with deadlocks when attempting to send too many API requests at the same time. Despite trying to resolve it for the past 8 hours, we still haven't figured out a fix.
Accomplishments that we're proud of
Being able to build a working prototype with the core features being functional.
What we learned
We obtained knowledge in using Genai APIs, full-stack development, as well as how to work in a team effectively.
What's next for InsightWallet
Fix the bugs that we have and polish on the features we are not yet finished with.
Built With
- cohere
- fastapi
- genai
- javascript
- llm
- mongodb
- python
- react
Log in or sign up for Devpost to join the conversation.