Gabriel Tseng
Hello! My name is Gabi. I am a Research Scientist at Ai2 on the OlmoEarth Team.
I completed a PhD at McGill / Mila under the supervision of Professor David Rolnick, investigating ways in which machine learning can help mitigate and adapt to climate change. This included working with Professor Hannah Kerner at NASA Harvest.
Land use is one of the most important environmental issues, and agriculture is the main driver of land use. My research seeks to better understand the Earth’s landcover using remote sensing data, with a focus on improving the machine learning algorithms used to make large scale agricultural landcover maps.
In addition to machine learning for remote sensing and agriculture, I have been lucky to contribute to:
- Technology for good, such as for sustainable and regenerative farming, or to increase energy access
- Interpretable machine learning using open source tools
I also love spending time outdoors, either by climbing or running in beautiful places.
In addition to the technical blog here, I also have a non technical blog, at oolongaloha.com
You can check out my CV here.
News
We released OlmoEarth v1.1. See our blog and tech report for more information.
OlmoEarth was accepted to CVPR! See you in Denver.
I’ll be discussing “Training and deploying pre-trained models for remote sensing data” at the ESA Φ-talk. Register here.
I defended my PhD!
We released OlmoEarth. See our webpage, tech report, platform blog and model blog for more information.
I’ll be in Rome speaking at the FAO’s Science and Innovation Forum (link).
I’m joining Ai2’s Earth System team as a Research Scientist!
I’m at ICML in Vancouver presenting Galileo. We’ll also be presenting 2 papers at the Terrabytes Workshop: our LFMC estimates using Galileo and WorldCereal’s integration of Presto.
I’m at the Living Planet Symposium in Vienna, where I’m talking about Presto and Galileo. In the same session, Christina Butsko will also talk about Presto’s deployment within the WorldCereal system. If you are also at LPS, come say hi!
Galileo is accepted to ICML! We develop a new self-supervised learning algorithm tailored to remote sensing, and use it to train a model which achieves state-of-the-art results across a diversity of remote sensing tasks (ranging from image segmentation to pixel-timeseries classification). Check out the code and paper.