October 21st, 2026 | 2:00 PM CEST
Live on Microsoft Teams.
On October 21st, the ESA Φ-Lab Collaborative Innovation Network will host a new Φ-talk. Details are below.
Meet the speaker
Eulalie Boucher is a Machine Learning Scientist at the European Center for Medium-Range Weather Forecasts (ECMWF). She joined ECMWF in May 2024 to develop machine learning-based weather forecasting directly from Earth observations. Prior to this, she studied mathematics and computer science. She completed a PhD at the Laboratoire d’Études du Rayonnement et de la Matière en Astrophysique et Atmosphères (LERMA) at the Paris Observatory, focusing on the use of deep learning techniques for atmospheric retrievals from the IASI (Infrared Atmospheric Sounding Interferometer) sounder.
Talk abstract
Eulalie will present the strategic and scientific motivation for developing an end-to-end machine-learning (ML) forecasting system for real-time operational use. The prototype system currently under development at ECMWF is based on AI-Direct Observation Prediction (AI-DOP), an approach that seeks to produce skillful medium-range weather forecasts directly from Earth system observations, including satellite radiances and in situ measurements, without using input from traditional numerical weather prediction (NWP).
The presentation will cover recent advances in AI-DOP modeling, progress in onboarding and curating observations, and the technical development of the prototype real-time system. It will also highlight key scientific and technical challenges in deploying an observation-driven ML forecasting system inside an operational environment.
She will also outline recent work to integrate observational data into Anemoi, the open-source ML framework developed jointly by ECMWF and meteorological agencies across Europe. This major update will allow Member States to build purely observation-driven or hybrid end-to-end forecasting systems, trained on both observations and (re)analysis data, for global and regional applications.
Register Here!