Thesis: Agent-Based Software Development-Specification-Driven Porting of a Sensor Tool

Summer Semester 2027 – limited to 3–6 months

Generative AI is fundamentally changing modern software development. Manual coding is increasingly giving way to agentic workflows in which AI agents independently generate, test, and iteratively improve code based on a specification. You will investigate the practical potential of this approach using a specific application: porting an existing sensor configuration tool to a new software platform. The focus of your thesis is to design and implement a specification-driven, agentic development workflow that enables you to carry out this porting process largely with AI support and evaluate the results.

YOUR TASKS:

• Familiarize yourself with the existing software and its functionality, and derive the specification of the current application, including through reverse engineering
• Design a specification-driven, agentic development workflow and select suitable tools and methods
• Implement a prototype of the workflow and use it to generate software for the target platform
• Port selected functions as a proof of concept and demonstrate the feasibility of the approach
• Evaluate the results using appropriate criteria, such as code quality, testability, development effort, and transferability, and derive recommendations for future use
• Document your approach and findings

YOUR PROFILE:

• Studying Computer Science, Software Engineering, Electrical Engineering, or a comparable field
• Good programming skills, ideally in Python and/or C; basic knowledge of web technologies such as Angular/TypeScript is an advantage
• Interest in generative AI, large language models, and agentic workflows
• Initial experience with AI-assisted software development is an advantage
• Independent, structured, and analytical approach to work, with an enthusiasm for exploring new topics

Contact: Sarah Disch