Projects
Personal projects repository covering the complete data cycle — from pipeline engineering to predictive modeling.
Data Engineering
Currency ETL Pipeline
Automated ETL pipeline that collects USD, EUR and BTC exchange rates in real time, transforms with Pandas and persists in PostgreSQL. Orchestrated with hourly Airflow, containerized with Docker and with CI/CD via GitHub Actions.
Data Science — Machine Learning
Decision Tree Classification
Decision Tree with Gini criterion for supervised classification, achieving 94% accuracy on test set. Includes feature importance, tree visualization and comparison with Naive Bayes.
Credit Score with Naive Bayes
Credit risk classification with Gaussian Naive Bayes — 98% accuracy and 99% macro recall on test set, with excellent generalization capability.
Real Estate Linear Regression
Simple and multiple regression for rent price prediction. Robust outlier treatment via IQR winsorization, multicollinearity analysis and overfitting diagnosis.
Customer Segmentation with K-Means
Unsupervised clustering of 200 mall customers by income, age and Spending Score. Choice of k validated by elbow, silhouette, Davies-Bouldin and Calinski-Harabasz, resulting in 5 segments (silhouette 0.555) with business strategies for each one. Reproducible environment with uv and Docker.
Titanic — Final Model Duel
A solution to Kaggle's Titanic competition built as an engineering project: leak-free pipelines, a leave-one-out group survival feature, 5×3 validation with paired comparison, tuning and ensembles. Hard voting with CV 0.8503 and 0.77751 on Kaggle, 65 tests and a reproducible Docker environment.
Data Science — Exploratory Analysis
E-commerce Dashboard
SQL pipeline that consolidates e-commerce transactions and customers via INNER JOIN, exports the consolidated dataset (296 records) and displays it in a self-contained HTML dashboard with 7 visualizations and interactive filters by gender, state and category.
Supermarket Sales Analysis
Full EDA on supermarket sales data identifying consumption patterns, seasonality and strategic insights. Descriptive statistical analysis with professional visualizations.
Software Engineering
EduConnect — Educational Management System
Complete Java system with 3-layer architecture, SOLID principles, inheritance, polymorphism and interfaces. Manages students, teachers, courses, classes and assessments.
Tower of Hanoi
Classic problem implementation in Python focused on software engineering: object orientation, SRP, YAML externalized configuration and robust CLI with argparse.
New projects coming soon
This portfolio is continuously updated. Follow GitHub for the latest releases.