I'm a Machine Learning Engineer focused on building end-to-end ML systems that work in the real world — not just in notebooks.
My work centers on imbalanced classification, production-ready model serving, time-series forecasting, and computer vision. I care about the full pipeline: from raw data to deployed inference.
Full-stack analytics system powered by ~94K real pharmacy transactions
- Built a Flask REST API with 9 endpoints and an interactive Chart.js dashboard
- Implemented Holt's Double Exponential Smoothing with auto-tuned parameters per product — 12-week demand forecasting with confidence intervals
- Built expiry risk tracking, staff performance comparison, and return rate analysis
- Accessible from any device on the local network with zero cloud dependency
Imbalanced binary classification on real credit card transaction data (~0.17% fraud rate)
- Evaluated Logistic Regression, Random Forest, and XGBoost using PR-AUC and cost-based metrics
- Reduced false positive alerts by ~94% while maintaining ~86% fraud recall
- Served the final model via a production-style FastAPI inference API with strict feature schema enforcement
CNN-based recognition system trained on 123K+ samples
- Applied image preprocessing, data augmentation, and model tuning
- Prepared trained models for inference and future deployment
End-to-end churn prediction pipeline on telecom data (~8K rows)
- Applied SMOTE balancing, feature engineering, and classification — 87% accuracy
- Delivered insights via SQL Server queries and Power BI dashboard
ML: Classification · Imbalanced Learning · Threshold Optimization · Time-Series Forecasting · Model Evaluation (PR-AUC)
Models: XGBoost · Random Forest · Logistic Regression · CNN
Stack: Python · Flask · FastAPI · Pandas · NumPy · TensorFlow · Scikit-learn · SQL · Power BI · Git