Thesis work within Embedded Software and AI
Agent-Based Generation of Release Documentation
We are looking for bachelor or master’s students in software engineering, computer science, artificial intelligence, or a related field who are interested in completing their thesis project within our embedded systems organization.
Release documentation is traditionally produced by engineers who review changes since the previous software release, identify what matters to users, and translate technical implementation details into clear descriptions of changed behavior, configuration, and limitations. This requires substantial contextual understanding and is difficult to automate with conventional rule-based methods.
This thesis explores whether an AI-based agent can generate reliable release documentation by combining evidence from a software repository with a document template and project-specific instructions for terminology, style, tone, and exclusions. The solution will be evaluated as part of an existing release process, with particular attention to documentation quality, traceability, robustness, and computational cost.
•
Design, implement, integrate, and evaluate an AI-based agent that automatically generates human-readable release documentation for a software product.
•
Investigate how much repository context is required to produce reliable documentation without excessive token consumption or loss of relevant information.
•
Enable documentation for earlier releases to be reconstructed from repository history, including patches and re-releases.
•
Evaluate whether the generated documentation is comparable in practical usefulness to documentation written by an experienced engineer, while clearly exposing uncertainty instead of inventing unsupported product knowledge.