AI Engineering

While this places AI Engineering clearly as a sub-discipline of software engineering, these systems and components are often different enough from traditional software systems that new and refined approaches are needed. For example, training and evaluation phases are needed as AI-enabled systems learn from data, adapt to changing contexts, and produce new outcomes using statistical methods. Requirements on such systems are often of a different nature, either implicitly defined through data or requiring a more statistical approach to describing both the expected behavior and tolerable behavioral variation. The fact that AI and ML systems and components also can learn and evolve after deployment poses additional demands both on how these systems are tested and how they are to be validated and certified/regulated. 

We see that modern language models like GPT or stable diffusion have profound impact on our society. They provide non-specialists with the possibility to use software technology for creating new products, author texts and create art. However, we still need to understand how to architect, develop and test such systems in a responsible and sustainable manner. 

Our research in AI Engineering aims to become a long-term, globally leading center for this type of research. Given the importance and impact of AI and ML technologies that we already see in society, it’s clear that these technologies are not going away but rather will just increase in importance. We are thus building knowledge on how to support society and industry in engineering digital systems based on AI and ML technologies.

Key Topics

• Engineering of AI-based products
• System and software architecture for AI-enabledbased systems
• Testing
• Requirement engineering
• Data pipelines and processing
• Development of conversational AI systems
• Cognitive Software Architectures based on Generative Language Models