AI & ML interests

Earth Virtual Expert (EVE) is a project by the European Space Agency (ESA) and Φ-lab (Phi Lab) to develop a Large Language Model AI expert from an open-source LLM, by fine-tuning it on specific and crafted documents related to Earth observation and Earth science, capable of several downstream tasks like QA and summarization, and with enhanced Retrieval Augmented Generation.

Recent Activity

alexratrio  updated a Space 9 days ago
eve-esa/README
jinoooooooooo  updated a dataset about 2 months ago
eve-esa/model-identity-questions-1k
jinoooooooooo  published a dataset about 2 months ago
eve-esa/model-identity-questions-1k
View all activity

EVE — Earth Virtual Expert

EVE Avatar

Earth Virtual Expert (EVE)
An open science initiative funded by the European Space Agency’s Φ-lab, developed by Pi School in collaboration with Imperative Space and Mistral AI, aimed at advancing the use of Large Language Models (LLMs) for the Earth Observation (EO) and Earth Science (ES) community.

Overview

EVE is a domain-specialized AI system designed to support scientists, analysts, and decision-makers through natural-language interaction.

It provides expert assistance across tasks such as:

  • EO and ES knowledge search and summarisation
  • Scientific Q&A and policy brief generation
  • Document understanding and retrieval
  • Technical and educational content creation
  • Integration of private EO document collections via RAG

The project unites data curation, model development, evaluation, and deployment to deliver open, reusable tools for Europe’s EO ecosystem.

Chat Platform

EVE features an interactive chat platform that allows users to explore and query Earth Observation and Earth Science knowledge through natural dialogue.
The platform is currently under closed access — you can request early registration here.
It is targeted for public release in early 2026.

EVE Chat Platform Preview

Model Development

EVE-instruct is trained on domain-curated corpora by fine-tuning Mistral Small 3.2 24B, through continued pretraining and instruction fine-tuning.

  • The model is optimized for factuality, reasoning, and integration with retrieval-augmented generation.
  • Training was carried out using EuroHPC Marenostrum, cloud environments, and Mistral infrastructure.
  • Evaluation targets both general domain and EO-specific benchmarks.

Licence

All EVE releases follow ESA and EU open-science principles, promoting transparency, reproducibility, and European AI sovereignty.
The project’s outputs—models, datasets, pipelines, and documentation—are shared under permissive open-source licences (Apache 2.0 / CC-BY).

Main Links

  • 🌐 Project Websiteeve.philab.esa.int
    Central hub for project updates, communication materials, and registration for early access to the EVE chat platform.

  • 🤗 Models & Data (Work in Progress)huggingface.co/eve-esa
    Repository for domain-specific model checkpoints, datasets, and evaluation benchmarks.

  • 💻 Codebase (Work in Progress)github.com/eve-esa
    Open-source tools, pipelines, and scripts developed within the EVE initiative.

Contributors

Pi School · Imperative Space · Mistral AI · European Space Agency (Φ-lab)

EVE Logo

models 0

None public yet