From September 22–25, 2025, this year’s NHR Conference took place in Göttingen, Germany. It brought together researchers, practitioners, and infrastructure experts from across Germany. The focus was on high-performance computing (HPC), AI, and their role in research and education. Various colleagues from ScaDS.AI Dresden/Leipzig presented their projects.

Presentations at the Conference

• Adaptive AI Mentoring through Multimodal Emotion Recognition and Real-Time Avatar Interaction
  Gitanjali Wadhwa, Oral Presentation
• Evaluating Prompt Architectures for AI-Driven Academic Mentorship: An HPC-Enabled Multi-LLM Approach
  Syed Hur Abbas, Oral Presentation
• AI Avatars in Higher Education: Sociotechnical Design and the Infrastructures of High-Performance Personalization
  Mana-Teresa Donner, Poster Presentation
• Mitigating I/O Bottlenecks in Training Deep Learning Models on HPC Clusters Using Data Sharding
  Ilya Burenko, Poster Presentation
• Automatic Annotation of Eyebrow Movement in a Modular Pipleline for Multimodal Corpora
  Ilya Burenko and Prof. Peter Uhrig, Poster Presentation

Outcomes

Attending the conference gave our team visibility and opportunities to engage with the broader research community. The junior research group “Situating AI-based Mentoring” in particular profitted from meaningful discussions and were able to generate valuable feedback on methods, infrastructure, and ethical considerations that will guide the next stages of their work.

The conference also opened doors for collaboration. A highlight was meeting a keynote speaker from the Zuse Institute in Berlin. This conversation has already led to plans for a joint project that involves university students, school students, and teachers. The goal is to create opportunities for cross-level exchange and explore new research approaches together.

The insights and connections from Göttingen are already shaping future work. The group will refine their systems and incorporate the feedback they received. Furthermore, they will advance the planned joint project with the Zuse Institute Berlin.