THE EUROPEAN SPACE AGENCY
THE EUROPEAN SPACE AGENCY
May 27, 2025
The EVE team recently participated in the NASA-ESA International Workshop on Geospatial AI Foundation Models at ESA-ESRIN. Representing the project were Pi School Deep Learning Scientists Àlex R. Atrio, Marcello Politi,Antonio Lopez, and Cristiano de Nobili.
EVE, still under development, was featured in early demonstrations—from dataset creation to training techniques, showing how LLMs and Retrieval-Augmented Generation (RAG) can make EO data more accessible for both scientists and the public.
Àlex R. Atrio delivered an oral presentation during the session Agent AI, Digital Assistant, Large Language Model presenting the project as a whole, its motivation, objectives, and overall pipeline.
During the deep dive session, Marcello Politi delivered a presentation on EVE’s use of domain-specific LLMs for semantic search, complex question answering, and scientific literature summarisation, tools that streamline research workflows.
A hands-on session co-led by Antonio Lopez and Muthukumaran Ramasubramanian (INDUS), demonstrated a practical RAG pipeline, highlighting the increasing synergy between AI and EO.
The workshop addressed the challenges and opportunities of adapting AI foundation models for Earth sciences, covering topics such as ethical AI use, geospatial data adaptation, and commercial applications. The event drew a global audience, both in person and online, reflecting the strong international interest in advancing AI for Earth science.
As we mentioned in the last newsletter, we will soon be launching the EVE pilot programme, an opportunity for you to test-drive the latest advancements in AI, designed specifically for EO and Earth Science. This is your chance to help shape a tool that could redefine how our community works with data and insights.
As part of the EVE pilot, you’ll gain early access to the cutting-edge Mistral 3.1 24B model, fine-tuned for EO and Earth Science applications. You’ll be among the first to experience new features, including an intuitive chat interface and seamless session memory, making your interactions smoother and more productive.
EVE is equipped with a sophisticated RAG system, giving you direct access to a wealth of high-quality resources, content from the ESA portal and the latest peer-reviewed articles. This means your queries will be answered with the most relevant, up-to-date information available.
During the pilot, you’ll interact with EVE in both freeform and guided chat scenarios, allowing you to explore its capabilities in real-world contexts. Your feedback is invaluable and there will be a chance for participants to share your experiences through quick surveys and integrated feedback tools. There are also opportunities to participate in follow-up interviews and workshops.
You’ll also play a crucial role in helping us spot and address any inaccuracies or inconsistencies, and you’ll be asked to rate EVE’s performance—especially its helpfulness, accuracy, and relevance for EO-specific tasks.
The EVE pilot runs from September 15 to October 15. Don’t miss your chance to be at the forefront of innovation—sign up today!
EVE at Living Planet Symposium 2025
Join us at the Living Planet Symposium in Vienna (23–27 June 2025) for two sessions on the EVE project
Hands-On Session: 22.06.2025 15 :30
Experience hands-on learning with EVE. Participants will gain insights into effective prompting strategies, learn how to use and build RAG databases, and discover ways to adapt EVE for specialised tasks. The session will also cover integration of EVE with external EO software tools, and explore key topics including domain adaptation, fine-tuning, best practices for retrieval and generation, mitigation of hallucinations, and evaluation metrics. This is a unique opportunity to deepen your expertise in leveraging AI for Earth science and EO applications.
Oral Presentation: 26.06.2025 08 :30 am
Join us for this presentation to discover the latest advancements in the use of LLMs across EO, climate science, and related disciplines. Learn how open-source LLMs are accelerating research, streamlining literature reviews, generating content, and supporting software development in the EO community. Gain practical insights into building digital assistants that enable natural language interaction, enhance data accessibility, and seamlessly integrate with third-party tools. Explore cutting edge techniques such as in-context learning, instruction fine-tuning, RAG, and prompt engineering to create intelligent agents, even when working with limited data.
Find more information about the Living Planet Symposium and the latest programme here
If you would like to meet any of the team at LPS please contact us here
Thank you for reading the second edition of the EVE newsletter, we appreciate your time and we will continue to provide updates as the project progresses. To stay informed and engaged, follow our social media channels and visit our website regularly for the latest news and don’t forget to sign up for the pilot programme!
