Investing Legends

Inspiration-

I am a Quant Finance student, and I have spent a significant amount of time studying the philosophies of the world's most successful investors. What really inspired me was how perfectly these strategies tied into the statistics and financial theories I’ve been learning in college. This project grew out of a genuine passion for the many hours spent reading financial reports and analyzing market trends.

What it does-

This simulator pulls real-world market and news data from a selected timeframe to fuel a weekly investment cycle driven by 14 unique AI characters. Before each trading week, your hand-picked team analyzes specific data points tailored to their individual philosophies such as Warren Buffett seeking value through low P/E and high ROE, or the Sentiment Scout gauging financial news "vibes" to pitch investment ideas.

These agents must then convince the lead Portfolio Manager AI to execute the trades, allowing you to see which strategic combinations actually survive the market. Because each character is powered by a lightweight Gemini 2.5 model, the outcomes are dynamic and unpredictable, offering a front-row seat to authentic, personality-driven arguments that evolve differently every time you run the simulation.

How I built it-

The project was developed using a combination of Lovable, an AI-powered website builder and runner, and Gemini to assist with the complex coding logic. The frontend is built with TypeScript and React, allowing for real-time streaming of the AI agents' "debates" and dynamic performance charting.

Challenges I ran into-

The path was filled with technical hurdles, specifically related to stability. Managing the orchestration of up to 8 concurrent Gemini models led to frequent glitches and crashes early on. Furthermore, moving from Java to TypeScript was a steep learning curve; I had to learn the nuances of the language on the fly to debug issues that weren't immediately obvious. Finally, API costs were a significant factor, forcing me to switch to lighter-weight models. While this occasionally makes the AI less sharp, it proved that the logic is scalable for more powerful models in the future.

Accomplishments that I'm proud of-

I am incredibly happy with how the website turned out. I was worried the interface would be a cluttered mess, but the final dashboard effectively displays complex data and AI dialogue in a way that is easy to digest. While I was confident in the AI logic, seeing the visual "War Room" come together was a huge win.

What I learned-

Through this project, I gained hands-on experience with TypeScript and modern web development. I also deepened my understanding of API orchestration, managing AI workflows, and how to structure multi-agent systems to solve complex financial problems.

What's next for Investing Legends-

The next step is to flush out the creation with a higher focus on "institutional grade" investing AI. While this version proves that agents can discuss and invest based on specific criteria, the future goal is to refine their decision-making capabilities to match the complexity of professional quantitative trading.

Built With

• google-gemini-api
• live-quotes
• lovable
• news-scraping
• postcss
• react
• react-query
• react-router
• recharts
• tailwind-css
• typescript
• vite
• vitest