Welcome to my GitHub profile. I'm a Backend & Data Infrastructure Engineer transitioning into Data Science, with a strong foundation in building the systems that data lives in — and now focused on extracting rigorous, statistically sound insights from it.
- Based in Santa Marta, Colombia.
- Pursuing an Especialización en Machine Learning Aplicado at Universidad del Magdalena.
- Background in backend engineering and data infrastructure: Go microservices, high-volume time-series pipelines, and OLAP optimization (ClickHouse, PostgreSQL, InfluxDB).
- Currently deepening my statistics and inference foundations — probability, hypothesis testing, experimental design, and statistical learning (ISLP).
- Experience running production game-server backend infrastructure: authentication (OIDC/SSO), metrics pipelines, and JVM tooling.
- Languages: Python, Go, SQL, Java, PHP
- Data Science & ML: Pandas, NumPy, Scikit-learn, Statistical Inference, Hypothesis Testing, Monte Carlo Simulation
- Databases: PostgreSQL, ClickHouse, MySQL, InfluxDB
- Visualization: Matplotlib, Seaborn, Grafana, Plotly
- Backend & Infra: FastAPI, Laravel, Docker, Linux, Nginx, OAuth2/OIDC
- Cloud: Railway, Hetzner, GCP APIs
-
Saber 11 Educational Outcomes Analysis — Santa Marta (2015–2025)
- Tools: Python, Pandas, Matplotlib, Seaborn, Parquet, Quarto
- Summary: End-to-end analysis of ~75K standardized test records across 22 ICFES datasets: modular ETL pipeline (loading, cleaning, enrichment), longitudinal trend analysis, hypothesis testing on public vs. private school gaps, and Spearman correlation analysis of socioeconomic factors against student performance.
-
Time-Series Pipeline Optimization
- Tools: ClickHouse, SQL
- Summary: Collapsed a 21.7M-row server metrics table to 458K rows via time-bucket aggregation with atomic table swaps — 47× reduction with zero downtime, enabling fast analytical queries over months of performance data.
-
Obbo — Reflection Proxy Library for Obfuscated Classes
- Tools: Java, Dynamic Proxies, Annotations
- Summary: A library for working with obfuscated or mapped classes through annotated proxy interfaces, with pluggable resolvers (JSON-based mapping) for translating symbolic names to runtime targets — useful for building version-independent integrations against obfuscated codebases.
-
Personal Finance Ledger from Email Parsing
- Tools: Go, Gmail API, PostgreSQL, Docker
- Summary: Reverse-engineered bank alert emails (6 formats, deduplication logic) into a structured transaction ledger — turning unstructured text into clean, analysis-ready financial data.
- Statistical Inference: OpenIntro Statistics + Think Stats / Think Bayes — verifying every concept with Monte Carlo simulation
- Statistical Learning: An Introduction to Statistical Learning (Python edition)
- Practice grounds: Kaggle Playground competitions, StrataScratch, and my own production datasets