QuantLab – ML-Driven Trading Simulator

Working on QuantLab has been a hands-on journey through the entire workflow of building a financial machine learning app from scratch using Python. Starting with the basics, I learned a lot about integrating external data sources, using yfinance to fetch historical stock data, and managing common data quirks like missing values, duplicated timestamps, and multi-index columns.

A big part of the project was structuring the data for predictive modeling. I implemented pipelines to generate features like percent and log returns, and set up a flexible forecast horizon so users can experiment with different prediction intervals. I also got to design the UI and UX with Streamlit, making the app interactive and visually engaging with Plotly charts, sidebar options, and real-time metrics.

On the machine learning side, I deepened my understanding of supervised regression, using models like Linear Regression, Random Forest, and XGBoost. I handled train/test splits, tracked evaluation metrics (RMSE, MAE, R²), and made the training process interactive, letting users tweak hyperparameters or swap models on the fly. Managing session state for user actions and model results was a practical lesson in building robust web applications.

Perhaps the most exciting part was building a simple trading strategy simulator. I learned how to translate model predictions into trading signals, apply realistic transaction costs, and compute key performance stats like Sharpe ratio and max drawdown. Backtesting strategies on out-of-sample data and visualizing equity curves made the ML results tangible and actionable.