Final Thesis: Design and Implementation of an AI-Enhanced Multi-Stage Testing Architecture for Scalable Software Quality Assurance

Abstract: This thesis investigates the design and prototypical implementation of an AI-enhanced multi-stage testing architecture to support scalable software quality assurance. The starting point is the existing test landscape of SCA Tool, which already comprises different testing levels but provides only limited support for traceable links to requirements, systematic identification of test gaps, and consistent evaluation of test results. The developed architecture combines standardized context collection with specialized agents for Requirements Review, Governance, AI-assisted assessment, and Test Generation. Pull request information, linked requirements, changed files, test artifacts, and traceability information are processed in a structured form. Particular emphasis is placed on the controlled use of AI: language models support analysis and the generation of test suggestions, while their outputs are constrained through deterministic checks, structured artifacts, evidence references, and human-in-the-loop mechanisms. The evaluation is based on a demonstration scenario and examines the extent to which the prototype can produce traceable findings, test suggestions, CI-compatible results, and reviewable artifacts. The thesis thus presents a prototypical approach to extensible, controlled, and artifact-based test support.

Keywords: AI-Assisted Software Testing, Software Quality Assurance, Multi-Stage Testing, Test Generation, Human-in-the-Loop

PDF: Bachelor Thesis

Reference: Roman Gailler. Design and Implementation of an AI-Enhanced Multi-Stage Testing Architecture for Scalable Software Quality Assurance. Master Thesis. Friedrich-Alexander-Universität Erlangen-Nürnberg: 2026.