Blog post: The Technical Debt of Earth Embedding Products
Louisa Diggs
Hi all,
Just sharing a recent post from the Cloud-Native Geospatial Forum (CNG) blog Isaac Corley, Senior Machine Learning Engineer at Wherobots.
Isaac writes that while foundation models have solved the challenge of training perabytes of imagery, the output is often unusable when you want to compare or stack different products. He shares lessons learned from their experience working with Earth embedding products, where the lack of standards required fixing patches for GDAL, Rasterio and TorchGeo just to get the data to load correctly.
The post explores what keeps breaking and why defining vector embedding standards is the next step for the ecosystem.
Read the full blog post here:
https://cloudnativegeo.org/blog/2026/02/the-technical-debt-of-earth-embedding-products
Best,
Louisa Diggs
Director of Operations