Systems and Knowledge Management for AI

This pillar studies the data, knowledge and computing infrastructures that support robust and trustworthy AI. We connect efficient learning systems with methods for representing information and reasoning about it.

Research directions

• Energy-efficient and privacy-preserving federated learning.
• Resource-aware optimization of large language models.
• Knowledge representation and reasoning.
• Semantic technologies, knowledge graphs and symbolic embeddings.
• Neuro-symbolic approaches and explainable AI.

Frugal and scalable intelligent systems

Making effective use of data, energy and computing resources is a scientific objective across the pillar. Distributed learning and resource-aware methods connect this objective with practical digital infrastructures.

Connecting learning and knowledge

Knowledge graphs, semantic technologies and symbolic reasoning provide structure for intelligent systems. These approaches complement statistical learning and support research in natural language processing, misinformation and interoperable systems.