Nadia Nahar is a Software Engineering PhD student at the Institute for Software Research in Carnegie Mellon University’s School of Computer Science. Her research asks a question that matters to every team shipping models: why do machine learning systems fail, and how can software engineering practices prevent it?

To answer it, she studies what makes ML different from traditional software, including uncertainty, complex data workflows, and monitoring challenges, together with the scope of hidden technical debt in ML systems. She has performed artifact analyses of around 300 open-source ML products, combining manual review with scripts over commits and code to identify common failure modes such as discontinued projects, unmet requirements, poor data, and deployment gaps.

Her work also spans requirements alignment, team structures, MLOps support, and documentation practices like Model Cards, Datasheets, factsheets, and checklists, alongside responsible AI topics such as explainability in healthcare and education and governance for product-centric fairness.