SentiTrade

Inspiration

We were inspired by platforms like AutoPilot and other assisted-trading tools that help users make better financial decisions. We wanted to go one step further by giving users control over which signals they trust, instead of relying on a single strategy.

What it does

SentiTrade fetches multiple data sources — historical stock prices, news sentiment, congressional trading activity, social media (Reddit) sentiment, market data, and pure math indicators. All signals are combined into a single system that produces clear, actionable trading insights.

How we built it

We created a set of data pipelines to gather all necessary information for SentiTrade. These pipelines fetch historical stock prices, news articles, Reddit posts, congressional trading activity, and market data. We process the text data using NLP models to extract sentiment, calculate mathematical and technical indicators, and feed all this into a GRU-based regression model to predict price movements. The outputs from all sources are combined into a single system that generates actionable trading signals for the user.

Challenges we ran into

Stock markets are inherently noisy and difficult to predict, even with multiple data sources. We faced challenges in selecting and finding the most reliable models, filtering NLP data to ensure it was truly relevant to each stock, and integrating new technologies efficiently within limited time.

Accomplishments that we're proud of

We delivered a fully functional, end-to-end platform that produces real-time, explainable trading signals. Users can choose which sources they trust most, making the system fully transparent. The project is production-ready and already useful.

What we learned

We learned how to integrate new platforms like GumLoop and Auth0, improved our understanding of trading strategies, applied NLP for sentiment prediction, and gained hands-on experience using GRU model for time-series forecasting.

What's next for SentiTrade

We plan to add more data sources, improve prediction accuracy, optimize algorithm speed, introduce risk management tools, and explore portfolio-level decision making.

Built With

• auth0
• geminiapi
• git
• github
• gumloop
• mondodb
• next.js
• python
• react
• tailwindcss
• tanstack
• typescript