Science. AI. HPC
The Mitiga recipe for boundary-pushing
climate risk intelligence
How we do it
Our understanding of climate risk is only as good as the quality of the data our models are trained on. So we use state-of-the art measurements of vegetation, weather dynamics, and topography. This lets us provide a comprehensive and accurate understanding of hazards in the physical world.
High-performance computing allows us to generate millions of scenarios every year to evaluate and refine the performance of our models at scale. By working with us, customers can access these HPC capabilities at new levels of affordability.
Our proprietary EarthScience AI™ technology leverages the latest machine learning transfer techniques to evaluate hazard and risk in data-scarce regions, based on learnings from data-rich ones. This lets us fill data gaps everywhere, making our models truly global by design.
Pioneering approaches
Mitiga started its journey as a spin-off from one of Europe's most powerful supercomputing centers. Our vision is to provide companies with the latest in science and technology. We believe in model transparency, disclosing unknowns to our customers and in the power of multi-disciplinary teams. Here are some of the approaches we have pioneered:
Combining physics-based models, probabilistic approaches and machine learning to create forward-looking climate risk models at various temporal scales.
Leveraging computer vision and deep learning to create a best-in-class, high-resolution, digital terrain model for flood analysis.
Representing the range of possible futures resulting from the interaction between human activity and climate change through our Multiple Futures Models.
Leveraging the combination of machine learning and HPC to generate better model inputs
at scale.
Widening access to highly sophisticated modelling techniques traditionally used by insurers. Probabilistic physics-based models using stochastic catalogues can now be applied to use cases in finance and asset management.
Mitigate climate risk differently
