Another Earth: synthetic satellite data to unlock AI for Earth observation

Training an AI model to detect deforestation, monitor infrastructure, or track land degradation from satellite imagery requires one resource that satellite programmes do not produce: labelled training data. Limited adoption of machine learning in Earth sciences has largely been attributed to this shortage, with non-representative spatial coverage identified as a source of bias in geospatial AI models.

Another Earth (Austria, founded 2024) was co-founded by Maya Pindeus and Felix Geremus in Vienna. The company builds a generative AI and 3D modelling engine that produces synthetic satellite imagery and geospatial datasets, automatically labelled and segmented, for training Earth observation AI models. The platform generates synthetic variations of satellite scenes, (e.g. edge cases, remote terrain, rare events), thus enabling organisations to train AI models without dependence on costly or incomplete real-world data collection.

The platform supports use cases including deforestation monitoring, biodiversity tracking, environmental risk simulation, and the assessment of carbon stocks - indicators that are increasingly tied to regulatory reporting obligations under frameworks such as the EU Deforestation Regulation.

In March 2026, Another Earth raised €3.5 million in seed funding from new investor Wake-Up Capital alongside existing investors Rockstart, Inovexus, and Stamco AG, with additional support from the Austrian Research Promotion Agency (FFG) and Austria Wirtschaftsservice (AWS). Commercial deployments are live with NovaTerra in Brazil and GeoTerra Image in Sub-Saharan Africa.

The company trains its multimodal diffusion models on EuroHPC infrastructure through AI Factory Austria, under the EuroHPC Joint Undertaking and Horizon Europe. As labelled training data has been identified as the primary bottleneck limiting AI adoption in Earth observation, the infrastructure Another Earth is building addresses a structural gap - without sufficient labelled data, the supervised AI models needed for applications such as deforestation detection and land degradation monitoring cannot be trained at scale.

Sources: Another Earth | AI Factory Austria | NASA IMPACT

Founders: Maya Pindeus, Felix Geremus