https://watch.eeg.cl.cam.ac.uk/videos/watch/42cea889-76a5-4f72-a542-cdd3d87a1241 While programs like ESA's Copernicus generate petabytes of satellite imagery, converting this data into useful models traditionally requires extensive labeling and bespoke training for every task. In this talk, we introduce the TESSERA Earth Observation Foundation Models, trained on AMD hardware and Vultr infrastructure. TESSERA compresses a year of temporal and spectral satellite signals into compact, precomputed embeddings at 10m resolution. Practitioners can leverage these embeddings to build accurate classifiers for land cover mapping, crop type identification, and change detection using a fraction of the standard data and compute. We will cover the model architecture, the large-scale training process, and practical applications for your own Earth Observation problems. Presented by Sadiq Jaffer at the RAISE 2026 Summit in Paris Tue, 21 Jul 2026 21:46:36 GMT https://validator.w3.org/feed/docs/rss2.html PeerTube - https://watch.eeg.cl.cam.ac.uk https://watch.eeg.cl.cam.ac.uk/client/assets/images/icons/icon-1500x1500.png https://watch.eeg.cl.cam.ac.uk/videos/watch/42cea889-76a5-4f72-a542-cdd3d87a1241 All rights reserved, unless otherwise specified in the terms specified at https://watch.eeg.cl.cam.ac.uk/about and potential licenses granted by each content's rightholder.