GenAI-powered software architecture

Architectural work involves a variety of people-centered activities, such as identifying architectural requirements, making design decisions, communicating with developers, and evaluating solution concepts. The quality of these activities is closely tied to the experience of the architects involved.

Today, generative AI can be used to complement human work in the field of software architecture, offering benefits to architects at all levels of experience: Less experienced architects can receive early feedback on their work and accelerate their learning curve; more experienced architects can evaluate different solution options more efficiently.

One concrete solution is Quasar, which was developed to analyze and evaluate architectural artifacts using generative AI, among other things, and to suggest improvements. We are also increasingly using the solution to jointly develop architectural solutions (AI pair engineering) and to take contextual information into account more closely. In particular, it is important to understand what delivers real added value and how significant that value is with Quasar. To this end, we conduct regular studies.

Learn about our research on AI-powered architectural design and discover how this technology can benefit your business.

• AI in Software Architecture: How LLMs Automate Quality Assurance[Blog in German]
• Shift-Left in Software Engineering: How AI Is Changing Roles[Video in German]
• Using LLMs to Evaluate Architecture Documents: Results from a Digital Marketplace Environment[Preprint]

Generative AI has opened the door to a new class of intelligent systems. Powered by agents based on language models (LLMs, SLMs), these systems can interact with users and other systems with a high degree of flexibility and perform tasks with a high degree of autonomy that traditionally require human effort to develop.

In the Department ACE, we develop technologies that enable systems to communicate with one another autonomously and reliably during runtime.

With the help of GRAIN (Generative AI for Interoperability), two systems can send and receive data in an unknown format. They can also immediately convert this data into their own internal format. This reduces development costs and increases the efficiency of data exchange.

Learn more about our research on intelligent information systems and use our solution to take data exchange to the next level in your organization as well. 

• Toward self-coding information systems [Preprint]
• Evaluating the effectiveness of LLM-based interoperability [Preprint]
• 2nd International Workshop of Software Architecture and Generative AI [Event]
• Implementing Generic APIs with Generative AI for Seamless Interoperability[Webinar]
• The Engineering of Smartness in Software-based Systems [Blog]
• The Pursuit of Smartness in Software-based Systems [Blog]