Enabling the Rapid Creation of Custom Predictive Models with GOES-16 and NEXRAD Data

NOAA's Big Data Project (BDP) has made several valuable datasets available on major cloud platforms. While a critical step, many organizations that would benefit from the data are not technologically capable of performing state-of-the-art machine learning. We propose to unlock the power of the data for those who can most benefit by developing technology that automatically trains deep learning models on GOES-16 and NEXRAD data, given a target objective. We have great expertise in performing large-scale deep learning on rich multimodal data (visual, quantitative, time-based, etc.), including for DARPA and NASA, and we propose to automate this expertise for the benefit of users across a range of industries—starting with agriculture—who can benefit from NOAA data. While cloud providers offer certain machine learning functionalities, there are no solutions in the marketplace targeted at fully unlocking the value of GOES-16 and NEXRAD data for those unfamiliar with big data technologies, let alone for nondevelopers.