February 18, 2026 14:00 CEST
Live on Microsoft Teams.
On 18 February at 14:00 CEST, the ESA Φ-Lab Collaborative Innovation Network will host a new Φ-talk. Details are below.
Meet the speaker
Gabriel Tseng is a research scientist at the Allen Institute for Artificial Intelligence (Ai2). He works on the OlmoEarth team, which is building a state-of-the-art artificial intelligence platform so that anyone can turn Earth data into timely, decision ready insights. Prior to working at Ai2, Gabi completed a PhD at McGill University / Mila investigating pre-training algorithms for remote sensing models with David Rolnick and Hannah Kerner. He has also worked closely with the machine learning team at NASA Harvest producing cropland and crop type maps around the world.
Talk abstract
Machine learning methods for satellite data have a range of societally relevant applications, but labels used to train models can be difficult or impossible to acquire. Self-supervision is a natural solution in settings with limited labeled data; when applied to remote sensing, these models must advantage of the characteristics of this data, including the temporal dimension (which is critical for many applications, such as monitoring crop growth) and availability of data from many complementary sensors (which can significantly improve a model’s predictive performance). In this talk, we discuss ongoing research to leverage the structure of remote sensing data to train performant models. In addition, we discuss the challenges and strategies associated with making these models accessible to the remote sensing community, and for deploying them at scale.