VITO launches SAIL: a lab for agentic AI grounded in physical reality
VITO is launching SAIL (Sustainable Agentic Intelligence Lab), a research initiative that develops agentic AI systems for applications in the physical world.
Most AI systems are trained on internet-based data. Such data is noisy, unstructured and lacks a connection to physical reality. SIAL addresses this fundamental limitation of current AI by combining agentic AI with validated scientific data from environmental monitoring, materials science and industrial systems. In doing so, we create AI that can reason and act within the operational logic of the physical world.
“The challenge is not to build ever larger models,” says Shane Ó Seasnáin, Director of Digitalisation and AI at VITO. “It is about building systems that are grounded in the real physical world, taking into account soil characteristics, atmospheric dynamics, material properties and infrastructural constraints. That is what VITO has been working on for decades: measuring and modelling.”
Three technical pillars
SAIL is a pioneering laboratory that focuses on three core areas:
- Agentic AI, aimed at multi-agent systems that can autonomously reason, act and coordinate using open standards.
- Context Engineering, the strategic design and management of the information provided to AI systems.
- Scientific Machine Learning, the development of physics-informed models and digital twins that combine simulation with data-driven approaches.
The lab will produce architectural patterns and reusable software components, domain-specific benchmarks, evaluation frameworks for context-aware AI, and energy efficient approaches aligned with the principles of sustainable AI. Together, these pillars deliver practical tools and frameworks for customers and partners in Flanders and Europe. This ensures that innovation can be deployed more rapidly in industry, that governments can use AI faster and more efficiently for more effective public services, and that SMEs gain a realistic pathway to integrating AI into their operations.
Agentic access to Earth observation
SAIL’s first project extends the openEO API, an ESA-supported open-source interface for federated processing of Earth observation data, with agentic capabilities. VITO already makes a significant contribution tot the open EO ecosystem and carries out large-scale implementations on the Copernicus Data Space Ecosystem and the Terrascope platforms. Users will no longer need to determine which data are available and how it should be processed, as AI agents will interpret analytical requests on demand and autonomously assemble processing workflows based on available tools and datasets.
“When we have access to Earth observation data and time series, we can rely on objective monitoring of changes on our planet. This forms a solid basis for modelling the future. Today, the transformation of this data into usable knowledge, insights and predictions is still largely carried out by highly trained experts. By adding an agent layer to the European reference implementations of the openEO API, we lower the barrier to using these vast volumes of data, enabling their impact and integration into future business processes to be fully realised", says Dennis Clarijs, Head of Remote Sensing Services at VITO.
“SAIL is committed to open-source development and European technological sovereignty by creating reusable AI tools that connect VITO products and contribute to a European AI ecosystem independent of proprietary platforms. By linking digital technology with our scientific expertise, we help industry and governments respond and anticipate much faster and more effectively to what lies ahead", says Inge Neven, CEO of VITO.
