We research and develop practical, innovative solutions at the intersection of software architecture, data engineering, machine learning, and synthetic data generation.
Our research areas in detail:
Software Architecture:
The research area Software Architecture focuses on the structured analysis, design, and evaluation of architectures of complex software systems. It encompasses both technological and methodological questions as well as concrete application contexts in which architectural decisions pose particular challenges.
At the methodological level, we investigate concepts and patterns for modularization, interface design, architecture evaluation, and the handling of non-functional requirements. From a technological perspective, a key focus lies on architectures for data-driven systems, including the integration of machine learning components, data flow management, and technical MLOps strategies.
From an application perspective, we address architectural challenges within enterprise-wide IT landscapes (enterprise architecture) as well as in domain-specific contexts, for example in logistics, public administration, or data-intensive platforms. We are particularly interested in scenarios in which complexity, dynamic change, or interdisciplinarity require careful architectural design.
Data Generation and Data Synthesis:
Our research area Data Generation and Data Synthesis aims to automatically generate realistic datasets, to deliberately enrich them, or to reduce them according to specific requirements. We develop rule-based approaches that enable the simulation of structured datasets based on domain logic and business rules. In parallel, we investigate generative machine learning approaches for the synthesis of tabular, relational, and multimodal data.
A central goal is to support AI and machine learning projects even when access to original data is limited by using synthetic data — for example, to increase model robustness or to generate diverse test datasets for software development. Our methods enable both the realistic representation of statistical distributions (for machine learning use cases) and a semantically diverse representation of data (important for high-quality testing scenarios). The replacement of sensitive original data with synthetic equivalents to ensure data protection and privacy is also part of our research focus.
From a technological perspective, we work with a broad range of modern generative models, including specialized GANs, VAEs, diffusion models, and transformer architectures. These are complemented by metrics for evaluating the quality, consistency, and privacy properties of the generated data. Our approaches are applied, for example, in test data synthesis or predictive maintenance and can be flexibly transferred to other domains.
Research Projects:
An overview of current and completed research projects can be found here.