Only three stickers today, but I got the coveted ICESat-2 🛰️ sticker!! Revisiting the NASA booths was worth it 😆

Also had a long chat with a new friend who's keen on helping out with PyGMT/GMT (need to transfer those ideas to GitHub issues 😅). Gonna try and catch up with some Pangeo folks tonight for Happy Hour 🥳 https://discourse.pangeo.io/t/agu-happy-hour-wednesday-evening/2992

#AGU22 #ICESat-2 #NASA #TOPS #PyGMT #Pangeo

More stickers again, triple the count from yesterday! Probably visited wayyy too many exhibition booths this afternoon, listening to some 'random' people from ESIP, NASA and other acronym organizations. The open science energy is going strong 🚀

Also shoutout to all the many Pangeo folks that have been spearheading the push towards open and collaborative science!

#AGU22 #NASA #TOPS #ESIP #OpenTopography #EarthAccess #OpenScapes #OpenScience #Pangeo

Recording is now up! ☯️ zen3geo: Guiding Earth Observation 🌏 data on its path to enlightenment!

https://www.youtube.com/watch?v=8uhOtQUTuDg

Yes, this is yet another geospatial machine learning library, with a UNIX-y design pattern twist that makes things lightweight but powerful. Keen to see more people use zen3geo for complex multi-modal data pipelines!

#gischat #GeoSpatial #DataPipe #Python #Pytorch #TorchData #Pangeo #EarthObservation #MachineLearning #GeoML

Slides for my ☯️ #zen3geo #Pangeo #ML talk on 🌏 #geospatial data pipelines are at https://hackmd.io/@weiji14/2022zen3geo. 📽️ Video will come once I get the recording!

- Walkthrough: https://zen3geo.readthedocs.io/en/v0.5.0/walkthrough.html
- Install: https://anaconda.org/conda-forge/zen3geo
- Discuss: https://github.com/weiji14/zen3geo/discussions

#gischat #DataScience #MachineLearning #STAC #CloudNativeGeoSpatial #EarthObservation #torchData #DataPipe #xarray #Python #FOSS #ML4EO

@edzer at the department we try to build infrastructure which is easy to deploy across scales.

What I find in lots of my work is that you want to leverage the frameworks on smaller problems as well (or those that are large in size, not in compute for #dataviz).

This means that it should work locally (small scale), as well as it would remotely (big services). #Pangeo + #OpenEo (interfacing) + #dataviz? will go a long way I think. Maybe someone as already setup a countainer?

I wonder what the place of the #fediverse and the #SmallWeb is in #science. I feel much tooling is gravitating toward big cloud platforms, which at times can mean vendor lock in (especially when considering geospatial platforms such as #GEE or similar). Although there are efficiency gains due to economies of scale, is there space for small modular geospatial platforms, and what would this look like? #Pangeo comes near, but not quite I feel. @edzer