CommonCents

Landing Page
Search Detail Page
Search Detail Page - Search History & Chatbot

Inspiration

Stock prices change quickly, and it can be hard for beginners to understand what’s going on—especially with today’s high volatility from international affairs. A tool that grounds stock analysis in recent news, explains things in plain language, and supports learning would be very helpful.

What it does

CommonCents lets you search for stocks by symbol or company name and view quantitative and qualitative analysis, price charts, ratings, and news. It also provides an AI chatbot for follow-up questions and saves search history for signed-in users.

How we built it

The frontend is a Next.js 14 app on Vercel with Supabase for auth and TypeScript. The backend is a FastAPI app exposed via Mangum on AWS Lambda, deployed with SAM and Docker. The project consists of quantitative and qualitative analysis. For quantitative, we get live stock price by using Alpaca API, and stock metrics by connecting Yahoo Finance API. We do first level analysis by calculating all the indicators and looking at their values. Then we built ML model to predict the average stock price of next 3 months for second level analysis. As for qualitative, we utilized LLM’s researching ability by making it searching headlines of news, targeting unquantifiable information such as geopolitics or macroeconomic events. Finally, we assigned AI three characters- Risk taker, Realist, and Conservative trader. Make three of them debate each other based on all the analysis we did, then generate final summary and overall rating of the stock of choice. The presentation of these things will be determined by the mode our users chooses. If beginners, terminology and technical explanations are prevented. If experts, we will include advanced information assuring great user experience.

Challenges we ran into

Deployment of the website
Building things within a limited time
Compromising different opinions within a team
Asking for help from other experts
To understand AI-generated code.
API connection
Quantitative finance knowledge
Finding ways of getting data

Accomplishments that we're proud of

Completing the project
Implement a fairly good design without a designer.

What we learned

How to deploy the website to the public
Teamwork
Building end-to-end product

What's next for Untitled

To establish the consistency of the rating score for the same stock, by building a more sophisticated system prompt.
To avoid hallucinations.
To consider more about the user experience.