FinFact | Devpost

Inspiration

As young adults, much of our financial advice comes from online sources, making it difficult to verify the accuracy of what we see. In a digital landscape filled with internet gurus flaunting Lamborghinis and selling courses that promise secret forex trading strategies, due diligence becomes increasingly challenging. After encountering a flood of dubious financial advice—ranging from misleading tips to outright scams—we wanted to create a tool that helps users critically assess these claims. Our goal was to provide a second voice that offers fact-based insights to give people better peace of mind about what they're seeing on the internet.

What it does

Our Chrome extension automatically detects user-generated financial advice on Reddit and Twitter—two major hubs for financial discussions, especially with the rise of crypto culture and options trading. When finance-related content is identified, the extension subjects it to AI-powered fact-checking. Users receive contextual analysis, fact-checked results, and proper reasoning supported by credible sources, helping them make informed decisions.

How we built it

We developed the Chrome extension using the Vite framework for the front-end and Flask for the back-end. The backend interfaces with a Cohere LLM, which receives scraped comments, tweets, and posts, analyze their content, and searches the web for reputable news articles. The AI then generates a detailed report that evaluates the accuracy of the post and explains its reasoning.

Challenges we ran into

One challenge we faced was ensuring that our backend consistently structured AI-generated responses in a uniform format. This was essential for delivering reliable fact-checking that would be easy for people from all sorts of financial backgrounds to understand. Additionally, we had to carefully design how the AI’s critiques would be presented to users—ensuring that the information was accessible and actionable without being intrusive or inconvenient.

Accomplishments that we're proud of

This project marked the first time our team successfully trained and deployed an AI model beyond simple API calls. We built a functional AI-driven system that actively benefits young adults and we're all very proud of this.

What we learned

We learned that more data doesn’t always lead to better results—whether in training an AI model or adding extra features. Thoughtful implementation and quality over quantity proved to be more effective.

What's next for FinFact

Moving forward, we plan to expand FinFact’s capabilities, aiming for a feature where a user can directly chat with our LLM to improve transparency, to ensure that our users can understand the AI's rationale through conversation. We would also add support for some of the other bigger sources of financial misinformation - Youtube and Instagram through video processing.