Inspiration
The backbone behind our inspiration is our passion for machine learning and AI. We wanted to utilize an LLM and various models to improve efficiency in finance. As engineers, we're more orientated toward tech companies. So solving problems regarding finance would be something new and also present new challenges.
What it does
Our project consists of 4 unique features that help users maintain a good financial standing. The first feature includes a budget advisor. The user can input a desired budget, and a model will project if the user exceeds their threshold in the near future. The second feature is detecting fraudulent transactions. During any abnormal transaction, the model will flag the user for fraud. Next is the chatbot feature where we implemented a smaller version of an LLM to give general financial advice. And lastly, we have a stock analyzer. This displays the top 10 - 15 stock options that are most profitable.
How we built it
We utilized various tools and frameworks to construct our idea. Google Colab was used primarily to train our models due to its free GPU usage. After gathering the models, we transferred our work from Google Colab to VS Code for future integration. We then used Github to improve efficiency and teamwork by synchronizing our work. Then Figma was utilized to develop the front end of the project, which would then be combined with the back end via VS Code.
Challenges we ran into
The first challenge we ran into was selecting the appropriate dataset for training one of our models. Initially, we faced various unbalanced datasets that didn't work with our use case. Another major hurdle was integration. It was very difficult to connect the back end with our front end. Some of the team members had issues with GitHub as well. That slowed down the process immensely and decreased efficiency.
Accomplishments that we're proud of
The main accomplishment would be implementing 4 different AI models within 24 hours. We didn't believe it was feasible to achieve more than 2 AI models. They all have unique attributes that, together, would help users make good financial decisions.
What we learned
One of the things we learned on the way was how to use React. It made web development easier to work with. Flask API was another useful tool we implemented. The process of connecting the back end and front end was a big learning curve. Axios and fetch were applied to make the process seamless.
What's next for your project
In the future, we are going to carefully consider the workload for each model implementation. We took more time trying to include as many features as possible and worrying about small details. Instead, we should focus more on the core aspect of our project to deliver a complete project.
Log in or sign up for Devpost to join the conversation.