BankBud

AI Financial Wellness Assistant that help users learn and improve their financial habits.

Comment 1

Inspiration

Recently, seemingly every part of the internet has decided to implement an AI feature and our group felt the need to create one that could help solve a regular and critical problem that can affect anyone. BankBud is a crucial tool for guidance in a complicated financial world.

What it does

BankBud gives important financial advice and guidance to people in all financial situations using generative AI. User text input is sent to the backend and carefully validated before a prompt is built to send to the OpenAI API.

How we built it

BankBud is built using React framework for the frontend, and Python Flask framework for the backend. LangChain was used to implement prompt templates for large language models. BankBud queries the OpenAI Api for the generative artificial intelligence responses.

Challenges we ran into

Building the entire frontend from scratch posed several challenges as the conversational flow between the user and the chat assistant has to be maintained (so the page can't be reloaded once the conversation is started). As for the backend, the prompt engineering took us some time to work out because, to obtain a precise and useful response, our bot requires certain information and those responses needed to be validated before prompt building. Our team also struggled to choose between 2 of the sponsor challenges because we loved the idea of working on either of them.

Accomplishments that we're proud of

We are extremely proud of completing the challenge of creating an entire AI assistant in just 24 hours!

What we learned

We learned generative AI prompt engineering using LangChain to communicate with the LLM. Additionally, our group as a whole learned several widely used fullstack technologies including: react and python flask.

What's next for BankBud

The future of BankBud: login functionality would allow users to save plans and keep them from having to re-enter financial situational information each time they wanted guidance.

Built With

Share this project:

Updates