Postdoc in AI‑Guided Code Repair Without Tests

University of Southern Denmark - SDU; Odense; On-site; DKK 450,000 - 600,000; Full time

Job summary

University of Southern Denmark (SDU) invites applications for a 2-year postdoctoral researcher position in the Software Engineering section of the Maersk Mc‑Kinney Moller Institute (MMMI). The successful candidate will contribute to the VILLUM Experiment project on test-free program repair, starting February 1, 2027.

The postdoc will design, implement, and evaluate methods that repair bugs using information already present in the code, without relying on test suites or specifications.

Qualifications

• PhD completed or near completion in Software Engineering, CS, or related field.
• Research experience in program analysis, mining software repositories, automated software engineering, testing, or repair.
• Interest in AI-generated/modified code reliability and empirical evaluation.
• Experience with empirical evaluation or bug benchmarks is advantageous.
• Proficiency in English with strong communication skills.
• Ability to work independently and within a research team.
• Strong research record or publications.

Responsibilities

• Design, implement, and evaluate a framework for test-free program repair.
• Conduct experiments on real-world bugs and AI-modified code.
• Publish results in journals and conferences.
• Collaborate with MMMI teams across languages.

Skills

Program analysis

Software repair

Mining software repos

Automated software engineering

Software testing

AI-generated code

English communication

Independent & collaborative work

Education

PhD in Software Engineering or Computer Science

Job description

University of Southern Denmark (SDU) invites applications for a 2-year postdoctoral researcher position in the Software Engineering section of the Maersk Mc‑Kinney Moller Institute (MMMI). The successful candidate will contribute to the VILLUM Experiment project on test-free program repair, starting February 1, 2027.

The postdoc will design, implement, and evaluate methods that repair bugs using information already present in the code, without relying on test suites or specifications.