Building AI-powered weather forecasting tools with Open Data on AWS

This article explores how Brightband uses AI and Open Data on AWS to build faster, more accessible weather forecasting solutions that can save lives and reduce economic losses.

• Brightband won 2024 Compute for Climate Fellowship for AI-powered weather forecasting
• Generates own weather analyses from raw observations, not just NOAA/ECMWF data
• ML models run on single GPU, produce forecasts in minutes at under $1 per forecast
• Forecasts competitive with or more accurate than NOAA and ECMWF models
• Uses NOAA GFS data from Registry of Open Data on AWS via S3
• Converts GRIB format data to efficient Zarr format for distributed computing
• Processes 5 years of historical data for model evaluation
• AWS QuickStart Jupyter notebook available for developers to access NOAA GFS data

Brightband democratizes weather forecasting by combining AI with cloud computing, enabling organizations of all sizes to generate accurate, location-specific forecasts sustainably.