Inspiration

We wanted to merge the idea of an AI chatbot with someone who could guide one to make better financial decisions.

What it does

Coin Assist is there to answer your questions! Ask away, however...remember it's not a real person! It's not real financial advice. Log in to your own account and see your chat history and ask new questions!

How we built it

From the ground up, it was built using a combination of Android Studio for the Android app, Firebase for managing users and their info, Cloudflare for hosting a Llama2 chatbot, and replit / flask (unnecessary but our school blocked our cloudflare worker url).
We started with a literal drawing board (fine, whiteboard) and drew up our vision. We planned out everything from the login screen to how we wanted the conversations to look!

Challenges we ran into

One of the first challenges was the school blocking our chatbot URL from Cloudflare! We later ran into issues listing out the conversations in the app, it's tricker than it looks! We also went back to UNC Charlotte where we ended up getting locked out of our working building for a little bit. And finally, the sad reality...our fair share of debugging errors!

Accomplishments that we're proud of

We were able to successfully manage our code and goals using Git, that is a big one. But also it was our first time trying to implement a chatbot of any sort and Cloudflare has made that super convenient! It's very satisfying seeing that we can make an app that works and it might be the best looking one we've made too!

What we learned

Planning is key! When we start with ideas early we can each knock out chunks that must be completed or looked into. That makes it easier to go back to the drawing board and helps mitigate what some might call sleep deprived chaos. Also big shout out to Cloudflare for making it very easy to deploy Llama2 and other AI solutions. Finally...prompting is super important!

What's next for Coin Assist

A more advanced model would be cool to have in the works. We think using OpenAI's assistant API would be very neat, especially since we could load assistants up with knowledge bases that could be tailored towards specific banks or specific products or ideas.

Share this project:

Updates