MSc Data Science candidate · Esp. Machine Learning
🌎 Spanish (native) · English (B2 · professional working proficiency)
Colombian Air Force captain (UH-60 Black Hawk copilot / UH-1 pilot) transitioning into data science. I discovered data while recovering from a stroke — building dashboards to track my own rehabilitation — and turned it into a career path. I bring cockpit discipline to data work: reproducible pipelines, statistical rigor, and clear communication.
| Project | What it does | Stack | Status | Links | |
|---|---|---|---|---|---|
| 🏋️ | rehab-strength | Three independent data sources → weekly snapshot with freshness badges; SHAP-explained models · sleep-quality classifier ROC-AUC 0.77 ± 0.15 (5-fold) — wide enough that it is directional, not diagnostic · one-click synthetic demo, so it opens with zero real health data | Python Streamlit Pandas |
 |
Live app » · Repo |
| 📰 | NLP-Fake-News-Colombia | BETO fine-tuned on 2,935 ColombiaCheck fact-checks (harvested from ClaimReview markup) · macro-F1 0.405 vs 0.386 TF-IDF baseline · Verdadero scores 0.00 and the model card says so — it is not a truth detector |
Transformers BETO Streamlit |
|
Live demo » · 🤗 Model · Repo |
| ⚡ | Forecast-Energy-StockChange-Colombia | 1.1M raw records → 101,256 gap-free hourly observations of Colombia's spot price since 2015, refreshed daily by a self-updating GitHub Actions pipeline. Naive settlement-version ordering left 47% of hours stale | Python Time series GitHub Actions |
 |
Repo |
| 🚁 | ASRS Rotorcraft NLP | MSc thesis — taxonomic coverage of NASA ASRS incident narratives in rotorcraft operations. Domain expertise from 1500+ flight hours applied to model auditing and disaggregated evaluation | Python NLP Transformers |
 |
private until defense |
From flight plans to data pipelines — precision, discipline and clear communication.