Work Package Leader — WP2
We lead WP2 and design ARGENTIC's reference architecture for distributed agentic intelligence, which becomes the project's open-source SDK. We also build the multi-agent collaboration and task delegation logic, the RAG 2.0 interfaces for secure knowledge retrieval, and the CI/CD pipelines for the model lifecycle.
Generative AI depends on centralised, compute-heavy infrastructure that raises energy use and latency and puts Europe's technological sovereignty at risk. These systems struggle with heterogeneous devices, decentralised data and on-device privacy. They also lack the flexibility for AI components that act autonomously but cooperatively under human oversight.
ARGENTIC runs explainable agents across IoT devices, edge, cloud and HPC. It combines lightweight language models, multi-agent coordination, federated learning, digital twins and human-in-the-loop governance, so every agent action is traceable and justifiable. Our reference architecture connects these layers into one sovereign platform, validated in healthcare, recruitment, tourism, marketing, energy, agriculture and virtual companions.
