Design and Implementation of an AI-based Agent to Inform Best Practices on Test Case Execution Routines

The rapid advancement of Large Language Models (LLMs) has introduced new opportunities in software and hardware development, particularly in assisting engineers with test case creation and automation. This thesis explores the application of LLMs in the automotive testing domain, focusing on the development and deployment of an AI-based assistant for test case generation. By leveraging fine-tuning and Retrieval-Augmented Generation (RAG), the proposed solution aims to enhance both the efficiency and semantic accuracy of test development workflows. The study begins with a comprehensive analysis of existing LLM architectures and adaptation methods to identify the most effective strategies for domain-specific tasks. Based on these insights, an AI-driven module is designed and implemented to support the requirements of automotive testing at Volvo. The system features a structured data processing pipeline and employs vector databases to encode semantic representations of test artifacts, enabling context-aware retrieval and generation. To evaluate the proposed approach, this work investigates several fine-tuning techniques for embedding models using a curated dataset derived from real-world requirements and test sequences. A comparative analysis of state-of-the-art open-source models demonstrates that embedding performance is highly sensitive to model selection, fine-tuning strategy, and data quality. These findings contribute to the emerging field of AI-assisted engineering by offering practical guidance on the deployment of LLMs in complex industrial environments.