Event
When an AI system answers a question, the quality of the answer does not only depend on the language model. It depends on which information the system retrieves – and how it evaluates that information.
Maximilian Schattauer presented at Conf42 Prompt Engineering 2025 on how retrievers and document ranking in AI systems can be systematically optimised. The talk built on our ongoing work at Perelyn: How do we ensure that the systems we develop for clients do not just deliver answers, but the right answers?
The question sounds simple. In practice, it is not. Documents carry different levels of relevance, contexts shift, and the order in which a model receives information influences the outcome. Without careful engineering, the result is systems that sound convincing but miss the point.
For us, prompt engineering is not an isolated topic. It is a component of system architecture – and therefore an engineering discipline, not a matter of style.
The recording of the talk is available on YouTube.
News
The proceedings from the "AI in Production" workshop at KI2025 have been published. Among the contributions is a paper by our team on the use of the Model Context Protocol (MCP) in industrial production environments.
News
Dominik Filipiak and Michael Banf are co-authors of a community paper on the 2025 Topological Deep Learning Challenge, published in the Proceedings of Machine Learning Research. Their contributions feed into TopoBench, an open benchmarking library for the research community.
