MSc Data Science graduate turning complex data into clear business decisions. I build predictive models and dashboards that help teams act faster and smarter.
Built a gradient boosting model to estimate claim risk for auto policies. Engineered features from policyholder data that improved premium pricing accuracy and reduced loss ratios.
Developed classification and clustering models to identify false alarms in emergency calls. Delivered actionable insights for deployment planning and resource optimization.
Benchmarked KNN, Random Forest, Logistic Regression, and SVM models. Identified BMD and Age as top predictive factors using feature importance analysis.
Trained sequence-to-sequence model on 45K LRS2 sentences using TPU. Implemented phoneme-viseme features achieving ~3% WER, outperforming baseline by 15+ percentage points.