Deep Sky

​​​

Every year, governments and private agencies collect petabytes worth of weather and climate observations. However, effectively incorporating these incomplete and noisy observations in traditional Numerical Weather Prediction (NWP) models remains challenging. 

​

Deep Learning has recently emerged as an alternative to traditional physics-based NWP models. At Red Sky AI we think Deep Learning represents the next big leap in weather forecasting, not only can a model be trained using the many petabytes of historical observations, greatly reducing biases, but a trained model is several orders of magnitude faster than a physics-based NWP model. 

​​

Here at Red Sky AI we are developing an operational weather forecast chain, comprising the creation of the initial condition, the global forecast, and client-specific regional forecasts.