Projects

An end-to-end data science system handling customer segmentation, churn prediction, and CLV forecasting using FastAPI and Scikit-learn. Built with a focus on OOP and sustainable architecture.

A comprehensive management system developed to practice modern Java web technologies. It features server-side rendering with Thymeleaf, dynamic project-specific task allocation, and an H2 in-memory database for rapid development. The architecture follows the MVC pattern, utilizing Spring Data JPA and Hibernate for efficient object-relational mapping.

A full-stack communication platform featuring instant messaging, user presence notifications, and a custom JSON message protocol for event handling. The project implements a centralized messaging architecture with Node.js, managing socket lifecycles (open, message, close) and ensuring secure connection validation. Deployed with a decoupled architecture on Render and Cloudflare Pages.

This project demonstrates a robust pipeline for credit card fraud detection, specifically addressing the challenge of imbalanced datasets through sub-sampling and hyperparameter tuning. It evaluates multiple models—including Logistic Regression, K-Nearest Neighbors, Support Vector Classifier, and Random Forests—using GridSearchCV for optimization. Performance is validated through cross-validation and detailed metrics like precision-recall and F1-score to ensure reliability.