Embed2Scale supports the next Geo-Embeddings Sprint in Zürich
The Embed2Scale project is pleased to support the next Geo-Embeddings Sprint, taking place on 27–28 October 2026 in Zürich, Switzerland. Hosted by IBM and organised together with Cloud-Native Geospatial (CNG), Clark University, and Planet, the event will bring together leading researchers, foundation model developers, data providers, and Earth Observation (EO) practitioners to advance the future of geospatial AI.
The second Geo-Embeddings Sprint builds on the success of the inaugural event held in Worcester (USA) in March 2026, where participants established the foundations for a growing international community around Earth Observation embeddings. Since then, important milestones have been achieved, including the launch of geoembeddings.org, the publication of embedding datasets using emerging community standards, and the development of common metadata conventions for EO embeddings.
As AI foundation models become increasingly important for analysing satellite data, the community is now focusing on one of its biggest challenges: how to evaluate, compare, and operationalise Earth Observation embeddings.
These objectives align closely with the goals of Embed2Scale, which develops AI-driven neural compression and embedding technologies to enable more efficient storage, access, and analysis of large-scale Earth Observation data.
During the two-day sprint, participants will address several key topics, including:
- Benchmarking Earth Observation embeddings across different applications;
- Defining fitness-for-use criteria for embedding products;
- Improving documentation and educational resources;
- Accelerating operational adoption beyond the research community;
- Identifying priorities for future standards and community collaboration.
The event also provides an excellent opportunity to strengthen collaboration between European research projects, industry, academia, and the wider open geospatial ecosystem.
Embed2Scale has actively contributed to the development of the geo-embeddings community through its research on neural compression, self-supervised learning, benchmarking methodologies, and AI-powered Earth Observation workflows. Participation in this sprint will further support the project’s mission of enabling scalable, interoperable, and sustainable Earth Observation data infrastructures.
Applications to participate are open until 30 August 2026, while the broader community is also encouraged to contribute through the ongoing Earth Observation Vector Embeddings Survey.
We look forward to contributing to the discussions and helping shape the next generation of AI-enabled Earth Observation. You can register here and the livestream will be offered on both LinkedIn and YouTube.