FinTrack

Home Page
Prompt + Code Output
Auth0 User Authentification

Inspiration

The idea for FinTrack was born out of the need to bridge the gap between novice investors and the complexities of algorithmic trading. Many people have ideas for trading strategies but lack the coding skills or the know-how to bring those strategies to life. FinTrack was created to empower users to design, test, and refine trading strategies without needing to write any code.

What it does

FinTrack takes user input describing a trading strategy (e.g., “buy when RSI is below 30, sell when above 70”) and generates a custom Backtrader strategy in Python. The platform provides a step-by-step breakdown of each part of the strategy code, allowing users to understand how the strategy works. Users can refine their strategies by adjusting parameters and receive clear explanations, making it an ideal tool for both beginners and experienced traders looking to fine-tune their trading approaches.

How we built it

FinTrack was built using a pre-trained DeepSeek R1 Distill Qwen 7B model, which we fine-tuned specifically for generating Python trading scripts. This fine-tuning ensured the model could accurately translate user input into clean, executable Backtrader strategies tailored for stock trading. The platform leverages React for a dynamic and responsive front end, allowing users to easily interact with the system and receive intuitive, personalized strategies. The backend focuses on providing detailed explanations for each part of the generated code, enabling both beginner and advanced traders to refine their strategies effectively. We designed the system to empower users to optimize their trading strategies, even without coding experience.

Challenges we ran into

One of the primary challenges was getting the Jupyter server up and running smoothly. Configuring the back-end environment to support both the AI model and the backtesting environment required several tweaks, and it took time to ensure everything was working without conflicts. We also faced issues fine-tuning the parameters for trading strategies. Identifying the right parameters for various indicators and optimizing them for real-world conditions proved to be a tough task, as markets can be highly volatile and complex. Additionally, creating a system that could consistently generate effective strategies for users required a lot of testing and iteration. On the front-end side, implementing interactive buttons and ensuring they triggered the correct actions, like generating strategies and explaining code, was tricky. We also had to work hard to improve the UI—making sure the platform was intuitive while maintaining its functionality for more experienced users. Ultimately, these challenges helped us refine the product and led to a better, more user-friendly platform.

Accomplishments that we're proud of

We’re particularly proud of integrating Auth0 for secure authentication, allowing users to safely log in and access their customized strategies. This streamlined access while ensuring user data privacy. We also successfully integrated an LLM (Large Language Model) into the front-end, enabling real-time interactions between users and the AI to generate and refine Python-based trading strategies on demand. The most exciting part is our achievement of 90% accuracy in the LLM’s output. By fine-tuning the model specifically for Python trading scripts, we were able to generate highly accurate and actionable code that users can implement directly into their own backtesting environments. This level of precision is a huge win, as it reduces the need for manual intervention and allows users—whether beginners or more experienced traders—to access reliable trading strategies effortlessly.

What we learned

We learned that combining AI with trading strategies opens up opportunities for education, innovation, and empowerment in finance. We also realized that helping users understand how their trading strategies work is just as important as giving them the tools to create them. Building a product that balances simplicity, flexibility, and advanced functionality requires constant iteration and a deep understanding of both finance and software development.

What's next for FinTrack

Next, we plan to enhance FinTrack with more advanced features like backtesting integration, where users can simulate their strategies with real market data. We also aim to build a community of traders who can share their strategies and learn from each other. Adding additional indicators and more customization options will help users tailor their strategies even further. Ultimately, we envision FinTrack becoming an essential platform for both novice and seasoned traders looking to refine their strategies and gain a deeper understanding of algorithmic trading.