PhD Stipend in Energy-Efficient Agentic AI: Measuring When AI Is Worth Its Energy Aalborg University in Denmark

This PhD stipend is hosted by the Department of Computer Science at Aalborg University and is embedded in the Distributed Sustainable Software and Systems (DSSS) research group within the Distributed, Embedded and Intelligent Systems (DEIS) section. The project addresses a timely question in AI systems: how much energy agentic AI actually consumes, when that cost is justified, and when a cheaper alternative would be better.

The PhD is part of the broader collaboration Energy, Sovereignty and the Responsible Use of Agentic AI, which connects technical measurement work with organisational and responsible-use perspectives. On the technical side, the student will investigate energy use in large language model and agentic-coding workflows, with a particular focus on successful task completion rather than isolated model calls. The research is experimental and centered on curated task sets, direct hardware measurements for local models, and validated estimates for API-accessed frontier models.

The project is strongly rooted in practical computing systems research. The student will work with energy-measurement testbeds and tooling from the GreenCode and MAST projects, use on-premise GPU/CPU infrastructure, and have access to AAU’s CLAAUDIA cloud and HPC services. Research questions include comparing runtime agent calls with generating deterministic software once, studying break-even points across models and frameworks, and evaluating how prompt caching, model cascades, routing, and distillation affect energy per successful task. The work also touches reproducibility and open science, with pre-registered protocols, released measurement harnesses, and open-source benchmarks.

The PhD student will be part of a collaborative environment that includes the Aalborg University Business School and its IKE group as well as the National Centre for AI in Society (CAISA). A second PhD student at the Business School is working on organisational decision-making around AI, and the two students are trained together as a paired cohort. The student will also join shared seminars, contribute to pilots with Danish public and private organisations, and, in the early phase, work with a postdoctoral researcher experienced in agentic software development.

Within the Department of Computer Science, the student will join a large research environment known for sustainable computing, embedded and cyber-physical systems, AI, and software-intensive systems, with strong industrial collaboration and a problem-based learning culture. The position is designed for a researcher who wants to combine careful experimentation, energy-aware computing, and AI system analysis in a highly collaborative setting.