Blog
Perspectives on how AI and data are transforming modern businesses. Discover practical approaches, implementation strategies, and technical insights from our team.
Why XGBoost’s Gradients Aren’t the Same as Neural Networks
This article explains the fundamental differences between how gradients are utilized in XGBoost compared to neural networks. While neural networks use gradient descent to continuously fine-tune internal weights, XGBoost applies gradients directly to the predictions to see how much the model is missing the mark. Instead of tweaking existing parameters, it highlights how XGBoost uses these calculation errors to stack entirely new decision trees that step-by-step correct previous mistakes. Ultimately, the piece clarifies that neural networks optimize within a "weight space," whereas gradient boosting models optimize within a "prediction or function space."
FXFlow: Automated Web Scraping Pipeline for CBSL Exchange Rates Using Playwright
The article explores how a fully automated pipeline can be built to scrape and store daily exchange rate data from the Central Bank of Sri Lanka. Using tools like Playwright and AWS services, it demonstrates how web data can be reliably extracted, processed, and delivered through a scalable, serverless workflow.