rasdaman newsletter 03/2026
AI-Cubes: Making AI Simpler, Safer, Faster
January 2026 - In its newest release, rasdaman introduces seamlessly integrated AI support through the groundbreaking AI-Cubes™ technology.
AI-based analytics on large spatio-temporal datasets is a powerful new tool - however, not so easy to handle, in particular on Big Earth Data. This gap is closed by the latest rasdaman release introducing AI-Cubes™ to the AI community. As a datacube platform, rasdaman does not offer built-in (read: hard-coded) models, but rather is open to any of the many models existing. Thus, AI-Cubes™ enable simpler, safer, and faster AI for all.
- AI gets simpler: Simply ask datacubes in your own language, and get an answer: "Using CoperniCUBE™ datacubes, look into the past 10 years and generate a map of photovoltaics effectiveness". This simplicity is in strong contrast to the common programming burden needing some 100+ lines of expert Python code for model inference. With AI-Cubes™, AI on datacubes is no longer reserved for highly-trained experts, but becomes a versatile tool easily handled and adapted also by non-experts.
- AI gets safer: Invocation of models undergoes checks on server side on ensure only valid calls are made. The underlying novel concept of server-based Model Fencing™ for enhanced model safety and accuracy is innovated and advanced in EU EFRE FAIRgeo, co-funded by the European Union and State of Bremen.
- AI gets faster: The built-in scalability boosters of databases can make AI significantly faster than conventional Python and other procedural code, while still allowing GPUs etc. to be used. Tests with rasdaman AI-Cubes™ have confirmed this convincingly.
Users can upload, run, and share ML models. To this end, rasdaman has been extended with user-friendly model management coming with a Web GUI frontend. A growing set of widely used models is supported, including BigEarthNet, NASA/IBM Prithvi, and RS-LLaVa. Model metadata enhance the STAC MLM (Machine Learning Model).
Notably, AI-Cubes™ fully benefit from existing rasdaman boosters such as location-transparent federation and distributed query processing.
Ultimately, AI-Cubes™ as introduced by rasdaman represent a milestone in democratizing Big AI Analytics.
contact: Prof. Dr. Peter Baumann