Medical Doctor combining clinical reasoning with Python, SQL, healthcare analytics, and clinical informatics to turn healthcare data into reproducible, decision-ready insights.

My portfolio focuses on hospital quality, clinical data analysis, interoperability, and responsible predictive modeling, with attention to data provenance, limitations, and clinical context.

Open to Healthcare Data Analytics, Clinical Analytics, Health Informatics, and Public Health Data opportunities in the UAE, Gulf region, and internationally.

Problem: Explore hospital-level readmission performance while keeping patient-level predictive research methodologically separate.

• Data: CMS HRRP FY 2026 — 18,330 hospital-condition records across 3,055 hospitals and six HRRP conditions
• Methods: data-quality analysis, hospital/condition/geographic comparisons, reproducible Python analytics, classification research, threshold analysis
• Deliverables: Live Dashboard · research simulator · model card · documented decision trail
• Stack: Python · Pandas · Scikit-learn · JavaScript · Healthcare BI
• Principle: hospital-level quality signals are never presented as individual patient-risk estimates

Problem: Move from structured cardiovascular data to reproducible exploratory analysis and leakage-aware baseline modeling.

• Data: 918-record processed heart-disease dataset with documented upstream provenance
• Methods: EDA, clinical-variable visualization, train-only preprocessing, stratified hold-out evaluation, cross-validation
• Models: Logistic Regression and Random Forest
• Result: Logistic Regression hold-out ROC-AUC 0.933 in the published analysis
• Deliverable: executed notebook with exported analytical figures

Problem: Bridge interoperable clinical resources and analytics-ready data while preserving clinical meaning.

• Standards: FHIR · HL7 v2 · ICD-10-CM · LOINC · UCUM
• Workflow: synthetic FHIR resources → Python validation → terminology/reference checks → analytics-ready CSV + JSON validation report
• Resources: Patient · Condition · Observation · MedicationRequest · Encounter · Bundle
• Privacy: fully synthetic patient scenario; no PHI

Clinical-data cleaning, EDA, statistical testing, effect-size analysis, multiple-testing correction, and automated notebook verification with GitHub Actions. Dataset provenance and redistribution status are explicitly documented rather than assumed.

Transparent model-artifact audit demonstrating integrity validation and responsible documentation when the original training dataset and complete training workflow are unavailable. Unsupported performance claims are intentionally excluded.

Healthcare Analytics: Clinical Data Analysis · Hospital Quality & Readmissions · Public Health Data · Data Quality · Clinical Interpretation
Data: Python · Pandas · NumPy · SQL · Excel
Visualization: Power BI · Matplotlib · Seaborn
Statistics: EDA · Hypothesis Testing · Effect Sizes · Multiple-Testing Correction
Machine Learning: Scikit-learn · Classification · Model Evaluation · Reproducible Pipelines
Health Informatics: FHIR · HL7 v2 · ICD-10-CM · LOINC · UCUM
Workflow: Git · GitHub · GitHub Actions · Jupyter · Google Colab

My clinical background helps me approach healthcare datasets by focusing on the clinical question, unit of analysis, data quality, interpretation, and consequences of model errors before headline performance.

I aim to bridge medicine and data through transparent, reproducible analysis understandable to both technical and healthcare stakeholders.

• Portfolio projects are educational/research demonstrations, not medical devices.
• Predictive models require external validation, calibration, subgroup assessment, workflow evaluation, and clinical governance before real-world use.
• Dataset provenance, licensing, missingness, and unit of analysis are core analytical considerations.

🌐 Healthcare Data Portfolio
📊 Kaggle
💼 LinkedIn
📧 natheer.ne@gmail.com
📍 Libya · Open to UAE, Gulf, Remote & International Opportunities