Selected Work
Featured Projects
Real problems. Real data. Real outcomes. Explore the case studies — or run a live model simulation.
Delivery Downtime Prediction & Route Optimization
Delivery Downtime & Route Optimization
Predicting logistics delivery delays using operational, driver, and environmental signals to reduce downtime and improve routing efficiency.
- Driver performance & route conditions identified as strongest delay signals.
- Built a production-aligned feature pipeline for real-world deployment.
- Established a robust baseline for delay prediction in logistics systems.
RegressionFeature EngineeringPythonPower BI
customer-churn · Python / ML
Customer Churn Analysis & Segmentation
An end-to-end churn prediction system combining RFM feature engineering with machine learning to flag at-risk customers before they leave.
- Inactivity & refund behaviour identified as the strongest churn signals.
- 3 actionable customer segments surfaced via K-Means clustering.
- High-recall model ensuring no at-risk customer goes undetected.
PythonPandasscikit-learnSeaborn
fraud-detection · Random Forest
Fraud Detection Command Centre
A behavioural ML pipeline identifying fraudulent transactions by analysing velocity patterns, device signals, and time anomalies — visualised in a Tableau command centre.
- Random Forest with strong precision-recall on heavily imbalanced data.
- High-frequency burst patterns flagged as the primary fraud signal.
- Tableau dashboard tracking fraud trends & estimated financial impact.
PythonRandom ForestSMOTETableau