Designed and implemented a secure, real-time digital voting system using Django, PostgreSQL, and Redis Channels. Features include anonymous voting, live WebSocket updates, admin dashboard, category filtering, and vote audit logging. Engineered for enterprise scalability and integrity.

Developed a production style Python CLI to convert raw CSV drops into analytics-ready Parquet files with schema validation, idempotent runs, and structured logging. Implemented partitioned storage for efficient queries n built observability features including dead-letter queue (DLQ) handling n JSON logs. Achieved ~99% test coverage with GitHub Actions CI/CD. Added SQL window-function exercises with DuckDB for analytics practice. Deployed locally n to Amazon S3 with PyArrow n moto-based testing.

Designed and developed a WordPress-based portfolio to showcase PhD research, machine learning projects, and technical writing. Customized the theme and layout, added dynamic project and blog sections using custom PHP/JavaScript, integrated resume download and contact workflows, and optimized the site for responsiveness, performance, and accessibility.

Designed and developed a WordPress-based portfolio to showcase PhD research, machine learning projects, and technical writing. Customized the theme and layout, added dynamic project and blog sections using custom PHP/JavaScript, integrated resume download and contact workflows, and optimized the site for responsiveness, performance, and accessibility.

Developed a production style Python CLI to convert raw CSV drops into analytics-ready Parquet files with schema validation, idempotent runs, and structured logging. Implemented partitioned storage for efficient queries n built observability features including dead-letter queue (DLQ) handling n JSON logs. Achieved ~99% test coverage with GitHub Actions CI/CD. Added SQL window-function exercises with DuckDB for analytics practice. Deployed locally n to Amazon S3 with PyArrow n moto-based testing.

Developed and deployed an end-to-end ML system for real-time credit card fraud detection. Handled extreme class imbalance using SMOTE, trained and evaluated multiple models, selecting Random Forest (AUC ≈ 0.9619) for deployment. Built a FastAPI backend, Streamlit frontend, and Dockerized application deployed on Render with integrated Swagger API documentation.