How I built PaleoData Explorer (the long version)
I started with a simple search. I wanted to see where a species I was reading about had been found, and that led me to the Paleobiology Database. It is a global archive of fossil occurrences maintained by researchers, and it holds millions of records going back half a billion years.
The catch is that it all comes through raw JSON. I'm a student, not a paleontologist. I didn't want to write queries to look at fossils. I wanted to click a name and see its story.
The first prototype took one evening. A Streamlit app that called the API for a single taxon and dumped the results in a table. Ugly, slow, and honestly amazing. I searched for a species I actually cared about and watched the data come back. That moment is why I kept going.
The fun part was the visualisations. The hard part was the data. PBDB records arrive in wildly inconsistent shapes. Ranks that disagree. Geological intervals spelled different ways. Coordinates in different conventions. Most of the build went into that pipeline, and it is invisible when you use the app, which is exactly how it should be.
For hosting I chose Hugging Face Spaces. The app needed a backend and I didn't want to run a server. Spaces runs everything on their infrastructure. I push the code and anyone in the world can open the app in a browser, free, no install.
The app is live now. Search Tyrannosaurus rex, or a humble trilobite, and see where it lived: https://huggingface.co/spaces/benjamintia/PaleoData-Explorer
Next on the list: curated explorations, embeddable maps for researchers, and smarter search. Half a billion years of records, and there is still plenty left to build.
Log in or sign up for Devpost to join the conversation.