Objective
CEREBRA-AI will develop a European general-purpose AI platform designed to be adaptive, explainable and trustworthy. The project combines four complementary capabilities: neuro-symbolic reasoning, continual learning, interaction with physical systems and digital twins, and built-in mechanisms for transparency, provenance, fairness, uncertainty management and regulatory compliance.
Unlike conventional AI models that rely primarily on statistical pattern recognition, CEREBRA-AI will combine machine learning with structured knowledge and logical reasoning. This will enable the platform to carry out verifiable multi-step reasoning, recognise uncertainty, adapt to changing data and operating conditions, and explain the basis for its results.
A continual-learning orchestrator will select and combine specialised models according to the task, available computing resources and operational constraints. The platform will also connect AI reasoning and learning to robots, sensors and high-fidelity digital twins, grounding its knowledge in observations and interactions with real and simulated environments.
The integrated platform will be validated through nine pilots in healthcare, mobility, maritime transport, manufacturing, social care, knowledge work, agriculture and coastal environmental management. The project aims to advance its technologies to TRL 5 while demonstrating improvements in performance, efficiency, safety and trustworthiness.
SINTEF's role in the project
SINTEF contributes expertise in knowledge graphs, semantic technologies, neuro-symbolic AI, uncertainty-aware reasoning, data provenance and trustworthy data pipelines.
A central part of SINTEF’s work is the development of CEREBRA-AI’s dynamic neuro-symbolic knowledge base. This service will combine graph-based representations of entities, relationships and constraints with vector-based representations used by modern AI models. It will support multimodal information, continuous updates, conflict resolution and the integration of domain ontologies and pilot data.
SINTEF also contributes to the project’s multimodal logical inference engine. The engine will combine symbolic rules and constraint solving with neural models, planning and uncertainty quantification. It will produce verifiable evidence such as reasoning traces and constraint checks, making AI-generated results more transparent and suitable for use in decision-support systems.
In the project’s trustworthy-by-design framework, SINTEF contributes to a privacy-preserving and tamper-evident provenance system. The system will document data lineage, model and configuration versions, AI prompts, tool calls, reasoning traces, confidence estimates and decisions. Here, SINTEF will contribute with competence and artefacts on automated These records will create machine-readable evidence bundles supporting auditing, reproducibility and AI Act-aligned compliance.
SINTEF further contributes to two transport and environmental pilots:
- Maritime trajectory analysis: SINTEF supports the integration of heterogeneous sensor data and AI-based trajectory inference for maritime anomaly detection and rule-compliant decision support.
- Coastal resilience in the Oslo Fjord: SINTEF contributes to an AI-assisted forecasting service that combines high-resolution environmental modelling, graph neural networks, neuro-symbolic policy constraints and continual learning. The pilot will support transparent “what-if” scenarios for coastal ecosystem management and will connect with the European Digital Twin Ocean and Destination Earth infrastructures.
Through these activities, the Smart Data Group helps ensure that CEREBRA-AI can connect data, structured knowledge, AI models and decision-support services in an interoperable, traceable and trustworthy way.