Researchers, doctoral students, and collaborators from leading Nordic universities gathered in Aarhus, Denmark, on March 20–21, 2025 for the two-day NUEI Project Seminar, hosted at Aarhus University. The seminar brought together participants from the University of Helsinki, KTH Royal Institute of Technology, Aarhus University, and NTNU to discuss emerging research directions in edge intelligence, AI systems, extended reality, and distributed computing.
The event opened with welcome remarks and introductions by Qi, Sasu, and Xiaoli, setting the stage for two intensive days of technical presentations, discussions, and collaborative planning within the Nordforsk NUEI initiative.
The first session focused on inference in foundational models and explored the challenges of deploying large AI models in resource-constrained environments. Presentations covered language model inference at the edge, distributed transformer inference, and efficient routing and cascading techniques for large language models. Speakers highlighted the growing need for scalable and energy-efficient AI infrastructures capable of supporting next-generation edge applications.
The seminar continued with a session on augmented reality and content delivery, addressing latency-sensitive and synchronization-critical applications for multi-user XR systems. Researchers presented novel approaches for semantic-aware compression, direct outlier detection on compressed data, and adaptive neural routing systems for configuration-free content delivery across distributed infrastructures.
Deep reinforcement learning applications formed another major theme of the seminar. Presentations explored AI-driven radio resource slicing for tactile internet applications and reinforcement learning techniques for autonomous racing systems, demonstrating the increasing convergence of AI, communication systems, and real-time control.
An extended group discussion session provided participants with the opportunity to identify future research directions and strengthen collaboration between students and institutions. Discussions focused on joint research opportunities, long-term planning, and future project activities within the NUEI network.
The second day focused on efficient inference in IoT and edge environments. Topics included continual inference for AIoT systems, hierarchical inference algorithms, batch hierarchical inference at the edge, and touch-augmented 3D Gaussian splatting for enhanced reconstruction techniques. These presentations highlighted ongoing efforts to optimize AI performance across distributed and heterogeneous computing infrastructures.
The seminar concluded with closing remarks and a shared lunch, reinforcing the collaborative spirit of the NUEI initiative and its commitment to advancing cutting-edge research across Nordic institutions.
The NUEI Project Seminar successfully provided a platform for knowledge exchange, interdisciplinary collaboration, and the development of future research partnerships in edge AI, intelligent systems, and next-generation digital infrastructures.