BuildPilot

We are supporting Monto Tech GmbH in the BuildPilot project in developing a system for automated software development and quality assurance using generative artificial intelligence. The goal is to enable small and medium-sized enterprises to create customized software solutions without their own development team.

The core of the project is the automatic generation of application code and associated tests using large language models (LLMs). The code is then analyzed using tools such as SonarQube and integrated into a continuous integration workflow. A feedback loop generates improved prompts based on the analysis reports to iteratively optimize code quality, test coverage, and security.

Particular attention is paid to the use of locally hosted open-weight models (e.g., Mistral, LLaMA) to ensure data privacy and data sovereignty. Research focuses on the development of novel prompt engineering strategies, the evaluation of test quality, and the integration of LLMs into traditional development processes.