Skip to content

juanjararPYBM/ibm-data-science-portfolio

Repository files navigation

👨‍⚕️ Juan Pablo Jaramillo Ormaza

Medical Doctor → Health Data Scientist | Bridging Clinical Expertise & Data Science

Python IBM Certificate Status LinkedIn Domain

📍 Medellín, Colombia  |  💼 Open to Health Data Science & Clinical Analytics Roles


🧬 About Me

Private practice physician with 13+ years of clinical experience — now transitioning into Health Data Science through the IBM Professional Certificate.

My practice had two core dimensions. In general private medicine, I ran my own clinic managing a broad patient population with direct, longitudinal relationships. In oncology support and palliative care, I served as a clinical bridge: helping patients and families understand complex diagnoses, navigate treatment decisions, and face end-of-life with clarity — often as a last resort when the system had nothing more to offer.

That means I spent 13 years making high-stakes decisions under uncertainty, with incomplete data, where the cost of a wrong call was irreversible. That experience is directly relevant to building clinical decision support systems. I understand what a false negative costs not as a metric, but as a human outcome. I know why model explainability is non-negotiable in healthcare. And I understand the gap between what an algorithm outputs and what a clinician actually needs to act on it.

Foundation: 3 modules of Python for Everybody (University of Michigan) · IBM Data Science Professional Certificate (in progress)


🚀 Featured Project — Cardiovascular Risk EWS

Early Warning Score (EWS) system for ischemic event detection | Full CRISP-DM pipeline

A production-grade cardiovascular risk stratification system built to clinical standards — not a textbook exercise. Designed for real-world deployment with HL7/FHIR integration, HIPAA compliance, and clinician-facing explainability.

Phase Notebook Status Key Deliverable
📋 Business Understanding Fase 1 ↗ ✅ Complete EWS strategy · stakeholder analysis · ROI $8.2M
🔍 Data Understanding Fase 2 ↗ ✅ Complete EDA · Spearman heatmap (D3.js) · feature ranking
🛠️ Data Preparation Fase 3 ↗ ✅ Complete Feature engineering (slope_x_oldpeak, log_oldpeak, vessels_bin) · 18 features · SMOTE
🤖 Modeling Fase 4 ↗ ✅ Complete GradientBoosting ganador · Recall=1.000 · AUC=0.997 · GridSearchCV 5-fold
📊 Evaluation Fase 5 ↗ ✅ Complete Test set sellado · Recall=1.000 · AUC=0.9996 · FN=0 · SHAP top predictor: slope
🚀 Deployment Fase 6 ↗ ✅ Complete EWS deployable · 3 niveles (Bajo/Medio/Alto) · umbrales p<0.30/0.65 · artefactos joblib

Dataset: Cardiovascular Disease — Kaggle · 1,000 clinical records · 18 features (11 original + 3 engineered + 4 OHE)

Key findings:

  • slope (ST segment pendiente) confirmed as primary predictor — validated by 6 SHAP studies 2023–2026
  • ECG trio slope + oldpeak + exerciseangia provides strongest class separation
  • 11.3% of women present ischemic risk with no classic symptoms → gender-specific thresholds required in Fase 3
  • Business case: ~$504K/year per institution · $8.2M across 16 institutions at 20% mortality reduction
  • Gradient Boosting logró Recall=1.000 en test set sellado (0 falsos negativos)

👉 Master Index — Full Project Navigation


📂 Additional Projects

Course 1 — Medical Appointment No-Show Prediction

Dataset: 110,000+ Brazilian medical appointments  |  Status: ✅ Complete
Applied data cleaning, EDA and statistical analysis to identify no-show predictors. 20% no-show rate aligned with my private practice experience in oncology, where missed chemotherapy appointments lead to measurable disease progression.
/LABS/course-1-medical-appointments


🏥 Clinical Domain Expertise

Area Detail
General Private Medicine 13+ years running own clinic, broad patient population, direct longitudinal relationships
Oncology Patient Support Clinical bridge between patients, families and treating physicians — translating complex diagnoses into actionable understanding
Palliative & End-of-Life Care Last-resort support for patients the system had discharged — high-stakes decisions, limited information, irreversible outcomes
Clinical Decision-Making 13 years of diagnostic reasoning under real uncertainty, with real consequences
Health Informatics HL7/FHIR standards, EHR workflows, clinical interoperability
Regulatory HIPAA · Ley 1581 Colombia · patient privacy protocols

🎓 Certifications

Certificate Institution Issued
IBM Data Science Professional Certificate (in progress) IBM / Coursera
Tools for Data Science IBM Feb 2026
What is Data Science? IBM Feb 2026
Using Python to Access Web Data University of Michigan Jan 2026
Python Data Structures University of Michigan Dec 2025
Programming for Everybody (Getting Started with Python) University of Michigan Dec 2025

IBM Certificate — Course Progress

Course Status
Course 1: What is Data Science? ✅ Complete
Course 2: Tools for Data Science ✅ Complete
Course 3: Data Science Methodology ✅ Complete
Course 4: Python for Data Science & AI 🔄 In Progress
Course 5–10 📋 Upcoming

🛠️ Technical Stack

Languages: Python · SQL
Libraries: Pandas · NumPy · Matplotlib · Seaborn · SciPy · Scikit-learn (in progress)
Visualization: Chart.js · D3.js · Matplotlib (dark theme) · Seaborn
Tools: Jupyter Notebook · Google Colab · Git / GitHub · VS Code · Anaconda
Domain Standards: HL7/FHIR · HIPAA · CRISP-DM
Next: Scikit-learn ML pipelines · SHAP explainability · MLFlow


📂 Repository Structure

ibm-data-science-portfolio/
├── proyecto_medico/
│   └── cardiorisk/                         # CardioRisk — Cardiovascular EWS (CRISP-DM completo)
│       ├── Fase1_Business_Understanding.ipynb  # ✅ Complete
│       ├── Fase2_Data_Understanding.ipynb      # ✅ Complete
│       ├── Fase3_Data_Preparation.ipynb        # ✅ Complete | 18 features (3 engineered)
│       ├── Fase4_Modeling.ipynb                # ✅ Complete | GradientBoosting Recall=1.0
│       ├── Fase5_Evaluation.ipynb              # ✅ Complete | AUC-ROC=0.9996, FN=0
│       ├── Fase6_Deployment.ipynb              # ✅ Complete | EWS deployable
│       ├── Master_Index.ipynb                  # Navigation hub
│       └── README.md                           # Project documentation
└── LABS/
    └── course-1-medical-appointments/          # No-show prediction (110K records) ✅

📫 Let's Connect

"After 13 years at the bedside, the highest-leverage decision I could make was to scale clinical insight through data." Life at the service of life