Interested in partnerships around Trustworthy Foundation Models

A3 Lab (Aarhus University) offers core expertise in Agentic AI and Test-Time Adaptation (TTA). We build privacy-preserving, self-adapting foundation models resilient to real-world data shifts. Seeking to lead AI robustness WPs in Cluster 4 consortia.

I am an Associate Professor at Aarhus University (Denmark) and the Director of the A3 Lab (Adaptive and Agentic AI). We are actively seeking to join a Horizon Europe Cluster 4 consortium as a core European partner, specifically to lead Work Packages related to AI Trustworthiness, Model Robustness, and Adaptive Deployment.

Our Core Expertise: Our core technology is highly horizontal, domain-agnostic, and designed to provide the underlying infrastructure for reliable AI. We specialize in:

• Agentic AI Frameworks: We build autonomous AI agents designed to monitor, query, and manage large-scale foundation models during real-world deployment.

• Test-Time Adaptation (TTA): The biggest bottleneck in AI deployment is Out-of-Distribution (OOD) failure. We develop TTA algorithms that allow models to adapt on-the-fly to new environments, hardware, or domain shifts without requiring access to the original, restricted training data.

• Privacy-Preserving & Compliant AI: By adapting at inference time, our frameworks guarantee robustness and fairness across different demographics and environments while maintaining strict data privacy, directly aligning with the requirements of the EU AI Act.

What We Bring to Your Consortium: If you are coordinating a Cluster 4 proposal (targeting 2026 or 2027 deadlines), the A3 Lab can act as the dedicated partner handling the "Trust and Safety" layer of your AI technology. Whether your proposal focuses on digital healthcare, industrial automation, or secure infrastructure, we provide the adaptation mechanisms that ensure your applied models survive real-world perturbations and cross-border deployments.