A convolutional neural network that classifies chest X-ray images as normal or pneumonia-positive, achieving clinical-grade accuracy using transfer learning.

• Achieved 95% accuracy and 0.92 ROC-AUC on the Kaggle chest X-ray dataset
• Applied transfer learning with VGG16 and ResNet architectures with fine-tuning
• Preprocessed 5,000+ images with normalization, augmentation, and class balancing to address dataset imbalance
• Evaluated with precision, recall, F1-score, and AUC to validate clinical reliability

A fully normalized relational database for Formula 1 race analytics, designed from scratch using dimensional modeling principles to support fast, scalable analytical queries.

• Designed star schema with fact and dimension tables, surrogate keys, and conformed dimensions
• Applied 3NF normalization and indexing — improved query performance by 30%
• Managed 10,000+ race statistics across drivers, teams, circuits, and seasons
• Built complex SQL aggregations to surface lap time trends, driver rankings, and team comparisons
• Defined naming conventions and entity relationships for scalability and consistency

End-to-end data science project analyzing 200,000+ global health records to identify key predictors of life expectancy and build a high-accuracy regression model.

• Performed comprehensive EDA on 200,000+ records spanning 15 years and 190+ countries
• Identified socioeconomic and healthcare variables most strongly correlated with life expectancy
• Achieved R²=0.85 using XGBoost and Random Forest with feature engineering and cross-validation
• Evaluated 5+ regression algorithms and selected optimal model based on RMSE and cross-validated R²
• Built Tableau visualizations to communicate public health insights to non-technical stakeholders