Development Seed

What if interacting with geospatial data was as simple as asking a question? Our latest blog post explores how Large Language Models (LLMs) can serve as intelligent orchestrators for geospatial analysis—translating human language into precise spatial queries while ensuring reliable, verifiable results. By combining AI-driven workflows with trusted data sources and human oversight, we’re designing interfaces that are intuitive, multilingual, and useful for decision-makers, regardless of their GIS background. (Blog link in comments)

TL&DR: Add HTML output to your STAC API with a simple Middleware: https://lnkd.in/eDqwp6iy Today, STAC and stac-fastapi are the "de facto" metadata schema and application. They are used by most of Satellite providers and big organizations like ESA, NASA... While stac-fastapi applications are quite simple and fairly easy to setup, they have one big issue: They return JSON or GeoJSON documents. This is nice when you are a computer but as a user I want to be able to visually understand what the application is returning (I do not have a JSON parser integrated into my brain yet). This usually mean that If you want to increase the usage of your API you'll have to create a frontend application (or use stac-broswer https://lnkd.in/eEJasFF6). But what if the stac-fastapi application itself could return HTML documents directly? We started working on this while writing our custom eoapi stack (eoapi-devseed) where we are trying to push the limit of what an Earth Observation Stack can be. The result is pretty nice and you can see it on https://stac.eoapi.dev But adding HTML output to the application meant we had to re-write a lot of code which we know will be hard to maintain (as soon as there will be change in the stac-fastapi code). So we had to think about something else: Middleware (https://lnkd.in/et6iBc-d). Because stac-fastapi applications should always return responses based on the STAC specification, adding a middleware that listen for those responses and create HTML output is fairly simple. It also mean we can mostly only focus on the HTML part and not have to care about the stac-fastapi code itself. The project is still a Work-In-Progress but we are looking for feedback/contributions, so please checkout the repo!

I'll be in Geneva tomorrow and Wednesday for #HNPW, where I'll be listening and learning from experts in the humanitarian sector about their geospatial data needs. ?? With funding cuts from the dismantling of #USAID , I imagine geospatial needs are high, will budgets for manual analysis, duplicate data and costly closed-sourced software is low. Open-sourced, cloud-optimized data could help. ?? Please let me know if you're also attending and would like to connect!

??????????: ?????? ?????????? ?????? ??????????-???????????? ?????????? ?????????????????????? ???????? We're excited to host our first-ever hands-on eoAPI workshop at FedGeoDay 2025. For the great price of $20, you will get a 3-hour workshop with two of our team members, Henry Rodman and Emma L Paz. Here's what to expect: Description: eoAPI makes massive earth observation (EO) data archives discoverable and interoperable—the fastest and easiest way to configure, customize, and deploy a full Earth Observation stack. In this workshop, you will learn about the interoperable components of eoAPI and how to use them for cataloging, visualizing, and analyzing Earth observation data. 1?? 30-minute introduction to eoAPI and its components: pgstac: database for housing STAC metadata stac-fastapi-pgstac: STAC API for making a catalog searchable titiler-pgstac: visualization engine for data in STAC collections tipg: visualization engine for vector data in a Postgres/PostGIS database 2?? 60-minute guided walk-through in an interactive environment 3?? 60 minutes for open discussion, feedback and shared learning Go to https://www.fedgeo.us/ to register for FedGeoDay and secure your spot in the workshop (space is limited).

? ???????????????????????? ???????????????????? ???????????????? ???????? ???????? Last month, we collaborated with folks from NSF NCAR - The National Center for Atmospheric Research and NVIDIA during an?OpenACC?event (https://lnkd.in/gNZSiH7C) to streamline the flow of data from Zarr stores to NVIDIA GPUs. This work brings us a step closer to?GPU-native data analysis and ML workflows and involves coordinated changes to various upstream libraries like zarr-python, kvikIO,?and?xarray. We hope to eventually bring these improvements to?cupy-xarray?and showcase an end-to-end example utilizing NVIDIA?DALI?pipelines and?nvCOMP?GPU decompression for faster reads of large datasets like ERA-5 on Zarr. Keep an eye on what is happening in this?GitHub repo:https://lnkd.in/g_juf2aM https://lnkd.in/gEM_6YBY Wei Ji Leong (@[email protected]) Max Jones

Join us on April 2 for a geospatial happy hour at the Development Seed office in Lisbon. This is happening at the same time as we're hosting the Spring #STAPI sprint, organized by our friends at Element 84. You can expect a lot of people to geek out on space and geospatial. (including my brilliant colleagues Peter Gadomski & Emma L Paz) Interested in joining? Please RSVP here: https://lu.ma/sd6h6tfz so we can stock the fridge accordingly.

I’m heading to Snowbird, Utah for the Cloud-Native Geospatial Forum (CNG) Conference 2025 from April 30 – May 2! I’ll be coordinating a series of talks and roundtable discussions around building resilient data infrastructures. Really looking forward to great conversations on topics like *where* should geospatial projects live, open data licenses and creating planetary scale data institutions. Hope to see you there and reach out if you want to get more involved in these discussions! Register here:?https://lnkd.in/deH4Rxwp

?????? ?????? ??????????-???????????? ???????? ?????????????? ?????? ?????? ?????? ????????? Tell us your stories in the comments. ?? ??????????-???????????? ???????? ???? ???????? ???????????????? ???????? ???????? This article by Ryan Abernathey, et al., may have been published a few years ago, but it's still highly relevant. Cloud-native data repositories are not just a convenience but necessary for scalable, accessible, and impactful scientific research. The traditional "download model" of data distribution struggles to keep up with today’s petabyte-scale datasets. Cloud-native approaches offer solutions that enhance: ? Performance & reliability ? Cost-effectiveness ? Collaboration & reproducibility ? Access & inclusion The Pangeo Community (https://www.pangeo.io/) and others have demonstrated how open-source, cloud-optimized data architectures can reveal new scientific possibilities. As we continue designing systems for global-scale data, these best practices still serve as a valuable blueprint. ?? Check out the full article here: https://lnkd.in/g_x_g6_U Pangeo Community has a wealth of resources to get you started and is a friendly community for cloud-native geo and earth sciences. Get started with Pangeo at one of the European Geosciences Union (EGU) 2025 sessions: https://lnkd.in/gpkbzWs4 or join a Showcase (virtual presentation) to dive deep into a particular topic: https://lnkd.in/gkiuPw4x

Interested in working with geospatial data but not sure where to start—or how to improve? Join us virtually at #HNPW, **March 19th (tomorrow) at 4 PM UTC**, for a session by Development Seed and MapAction on free and open-source tools that make humanitarian data work easier, cheaper, more accessible, and more collaborative—no coding expertise needed! We’ll cover: ?? The current state of humanitarian data workflows ?? STAC for discovering and accessing data ?? Jupyter Notebooks for analysis and automation ?? Cloud-optimized data for efficiency and accessibility ?? Register for free here: https://lnkd.in/eBqrMhTS John Crowley, Henry Rodman, Emmanuel Mathot, Gjore Milevski, Tarashish Mishra

?? ???????????????????? ???? Our latest batch of experimental tools and open-source contributions explores: ?? ?? STAC APIs with a lightweight, object-storage backend (Peter Gadomski) ?? AI-powered knowledge management (Leo Thomas) ?? Wildfire data resources for emergency response (Kevin Bullock Max Jones) ?? Fine-grained access control for STAC (Anthony Lukach Zac Deziel) ?? GPU-native geospatial analysis (Wei Ji L.) ?? ???????? ???? ????????????????