Yawen Zhang
Currently, I am a Lead ML Engineer at the Allen Institute for AI (Ai2), working on building AI models and systems from research to real-world deployment. Previously, I was a Research Scientist at Google X and Mineral.ai, where I developed production ML models for crop yield forecasting and conducted research on multimodal models and LLMs.
I graduated with a Ph.D. in Computer Science from the University of Colorado Boulder (advisor: Qin Lv), where I worked closely with domain scientists, including atmospheric, fisheries, and seismology experts from organizations such as CIRES, NOAA, and LLNL, to develop ML models for real-world scientific problems.
Outside of work, I enjoy hiking, swimming, and exploring new foods and recipes.
Recent News
JUL 2026 • Our abstract "EarthDiT: Probabilistic Sub-km Weather Forecast Downscaling with Sparse Targets and Earth Observation Conditioning" has been accepted as an Oral presentation at the Workshop on Machine Learning for Earth System Modelling. See you in Bonn, Germany! 🇩🇪
JUN 2026 • We released OlmoEarth v1.2, which fixes the positional-encoding artifacts in our outputs. 🧩 [Paper]
MAY 2026 • We released OlmoEarth v1.1, a new family of models that cuts compute costs by up to 3x! 🚀 [Ai2 Blog] [Paper] [Code]
MAR 2026 • Our tutorial "Getting Started with OlmoEarth: From Embeddings to Fine-tuning" has been accepted at the ICLR 2026 Workshop on Machine Learning for Remote Sensing!
FEB 2026 • Our paper OlmoEarth: Stable Latent Image Modeling for Multimodal Earth Observation has been accepted at CVPR 2026. See you in Denver! 🏔️
DEC 2025 • Gave a talk "OlmoEarth: From Foundation Model Research to an End-to-End EO Platform" at the CS4Env seminar at the University of Washington.
NOV 2025 • We released OlmoEarth! 🌍 [Ai2 Blog] [Paper] [Code] [Platform]
JUL 2025 • Two papers accepted at ICML 2025. See you in Vancouver, Canada! 🍁
- High-Resolution Live Fuel Moisture Content (LFMC) Maps for Wildfire Risk from Multimodal Earth Observation Data – TerraBytes: Towards Global Datasets and Models for Earth Observation Workshop
- Are Time Series Foundation Models Ready for Zero-Shot Forecasting? – Foundation Models for Structured Data Workshop
MAY 2025 • Presented "Full Lifecycle of EO-based AI Foundation Models: Lessons Learned from Real-World Deployment" at the first ESA-NASA International Workshop on AI Foundation Model for EO.
MAY 2025 • The Landsat vessel detection model has been successfully deployed in Skylight (now public! 🎉), and added to the paper Satellite Imagery and AI: A New Era in Ocean Conservation, from Research to Deployment and Impact (Version. 2.0).