Skip to content

Data Wranglers at Geo for Good Dublin

If you are in this room, you signed up for the Data Wranglers Hackathon at the Geo for Good Mini Summit, Dublin. Welcome to our Hackthon challenge focused on data accessibility in Earth Engine using a new experimental Earth Engine Dataset Explorer, built by the Science AI team in Google Research! You are a data wrangler, whether you are a programmer, educator, or enthusiast using Google Earth Engine to solve small and large problems. Join us as we leverage Gemini, catalogs, and LLM technology to innovate how we interact with Earth Engine datasets.

Read more about the project background here

During this Hackathon we will interact with Large Language Models (LLMs), have open discussions on challenges with data discovery, and work on making some questions we can ask the dataset explorer together, as well as how better to make the datasets more discoverable through best practices that we hope you, as the participants, will share. Whether these are insights from classrooms or in-the-weeds challenges of those writing production code, we believe there is something for everyone.

This Earth Engine Dataset Explorer is implemented as a jupyter notebook and implements a Gemini-backed GEE search tool designed to help users discover datasets within the Earth Engine catalog that are relevant to their geospatial analysis tasks. The core functionality includes:

  1. Loading and processing Earth Engine dataset metadata and pre-computed embeddings.
  2. Implementing a dataset search function that uses vector similarity to find relevant datasets based on user queries.
  3. Creating an interactive user interface that displays search results, LLM-generated dataset details, code samples, and map visualizations.