Carnegie Mellon University

Christian Kästner

Dr. Christian Kästner

Associate Professor; Director, Software Engineering Ph.D. program

Address
5000 Forbes Avenue
Pittsburgh, PA 15213

Bio

I am an associate professor in the School of Computer Science at Carnegie Mellon University. My current interests are in software engineering for software systems with ML components (or teaching software engineering to data scientists, "machine learning in production"), open-source sustainability, and software-supply-chain security.

I am generally interested in understanding the limits of modularity and complexity caused by variability in software systems, which naturally brings me to questions of quality assurance, interoperability, and feature interactions. My research combines rigorous empirical research with program analysis and tool building.

Research

Research Interests

  • Analysis & Assurance
  • Developer Tools
  • Privacy and Security
  • Software Data Analysis

Software Engineering for AI-Enabled Systems

We explore how different facets of software engineering change with the introduction of machine learning components in production systems, with an interest in interdisciplinary collaboration, quality assurance, system-level thinking, safety, and better data science tools: Capturing Software Engineering for AI-Enabled Systems · Interdisciplinary Collaboration in Engineering AI-Enabled Systems · Developer Tooling for Data Scientists

Sustainability and Fairness in Open Source

We study the dynamics of open source communities with a focus on unstanding and fostering fair and sustainable environments. Primarily with empirical research methods, we explore topics, such as open source culture, coordination, stress and disengagement, funding, and security: Sustainability and Fairness in Open Source · Collaboration and Coordination in Open Source · Adoption of Practices and Tooling

Quality Assurance for Highly-Configurable Software Systems

We explore approaches to scale quality assurance strategies, including parsing, type checking, data-flow analysis, and testing, to huge configuration spaces in order to find variability bugs and detect feature interactions: Variational Analysis · Analysis of Unpreprocessed C Code · Variational Type Checking and Data-Flow Analysis · Variational Execution (Testing) · Sampling · Feature Interactions · Variational Specifications · Assuring and Understanding Quality Attributes as Performance and Energy · Security

Publications

Christian KästnerMachine Learning in Production: From Models to Products. 2022. (a final copy will be published late 2024 or 2025 by MIT Press). [ httpbib ]

Chenyang YangYining HongGrace LewisTongshuang Wu, and Christian Kästner. What Is Wrong with My Model? Identifying Systematic Problems with Semantic Data Slicing. In Proceedings of the 39th IEEE/ACM International Conference on Automated Software Engineering (ASE), Los Alamitos, CA: IEEE Computer Society, November 2024. Acceptance rate: 26 % (155/587). [  bib ]

Menon AlkaZahra Abba OmarNadia NaharXenophon PapademetrisLynn Fiellin, and Christian Kästner. Lessons from Clinical Communications for AI Systems. In Proceedings of the AAAI Conference on AI, Ethics, and Society (AIES), October 2024. [ .pdfbib ]

Nadia NaharHaoran ZhangGrace LewisShurui Zhou, and Christian Kästner. The Product Beyond the Model -- An Empirical Study of Repositories of Open-Source ML Products. In Proceedings of the 47th International Conference on Software Engineering (ICSE), April 2025. [ .pdfbib ]

Courtney MillerMahmoud JahanshahiAudris MockusBogdan Vasilescu, and Christian Kästner. Understanding the Response to Open-Source Dependency Abandonment in the npm Ecosystem. In Proceedings of the 47th International Conference on Software Engineering (ICSE), April 2025. [ .pdfbib ]

Nadia NaharJenny RowlettMatthew BrayZahra Abba OmarXenophon PapademetrisMenon Alka, and Christian Kästner. Regulating Explainability in Machine Learning Applications -- Observations from a Policy Design Experiment. In Proceedings of the Conference on Fairness, Accountability, and Transparency (FAccT), pages 2101--2112, June 2024. [ .pdfdoibib ]