FUTURES AI

Inspiration

We wanted to see if an AI model could reason like a trader, making disciplined buy or sell decisions based on real data beyond mere price prediction.

What We Built

• Collected daily market data via Polygon API.

• Engineered 50+ features (RSI, MACD, volatility, etc.).

• Used Gemini to decide actions (BUY / SELL / HOLD).

• Simulated portfolio performance over 2024 using Python.

What I Learned

• Clean data matters more than complex models.

• Technical indicators capture trader psychology.

• LLMs can combine quantitative signals with qualitative reasoning.

Challenges

• LLM outputs were inconsistent → enforced structured prompts.

Outcome

• We found that using this AI system outperformed simply investing in QQQ.

• Moreover, I built an interactive system that users can operate directly to visualize trades and simulate investment outcomes.

Built With

• gemini
• massive
• python
• react-native