Ai2 has launched the OlmoEarth Platform, an infrastructure system designed to take Earth observation foundation models from fine-tuning through to large-scale inference. The platform addresses the engineering challenges of processing terabytes of multimodal satellite data across multiple providers and resolutions.
- The execution layer divides jobs into three hardware-specific stages: CPU-based data acquisition, GPU-based inference, and CPU-based postprocessing.
- A continent-scale wildfire risk map for North America utilized 19,600 CPUs and 994 GPUs in parallel, achieving a 155× speedup over serial compute.
- The system maintains its own metadata index to avoid overwhelming external STAC APIs, using windowed reads against cloud-optimized formats like COG or Zarr.
- Inference across continent-scale areas takes roughly a day at a cost of fractions of a penny per square kilometer.
The platform enables organizations without large engineering teams to monitor environmental changes such as deforestation and wildfire risk by handling the full lifecycle of data preparation, model execution, and output stitching.