Building reliable software for complex, real-world problems.

A production-minded runtime for building, integrating, and operating enterprise AI workflows with clear boundaries between application logic, model services, data, and operations.

What it demonstrates: Shows the ability to turn AI capabilities into a dependable platform: service design, orchestration, integrations, deployment, and operational thinking.

A production-oriented medical-imaging workflow platform for ingesting, converting, de-identifying, cataloging, and viewing DICOM studies through a modern web application.

What it demonstrates: Demonstrates healthcare interoperability, privacy-aware data workflows, asynchronous processing, and end-to-end product engineering across the stack.

A unified Python research toolkit for facial-image kinship verification, bringing classical baselines, metric-learning methods, deep models, and reproducible experiment presets into one workflow.

What it demonstrates: Highlights research-grade experimentation, model comparison, dataset handling, evaluation design, and the discipline required to make ML results reproducible.

A benchmark for studying the trade-off between privacy protection and downstream utility in ultrasound imaging, with an emphasis on systematic, comparable evaluation.

What it demonstrates: Connects responsible AI, medical imaging, privacy-preserving methods, and empirical benchmarking into a focused research workflow.