Description: One of the challenges faced by the National Airspace (NAS) stakeholders in general and the Air Navigation Service Provider (ANSP) i.e., Federal Aviation Administration (FAA) in particular is the effective maintenance of safety in presence increasing demand by manned traffic and with introduction of Unmanned Air System (UAS) in the near future. DAAS architecture addresses the overall requirement for ensuring safe operations in the NAS while embracing its complexity and leveraging the advancements in machine learning through one simple architecture. The tools developed using DAAS will provide invaluable support to NASA and industry researchers in identifying, diagnosing and discovering the impacts of NextGen technologies on NAS safety and efficiency. Machine learning models have shown great promise in identifying outliers, predicting anomalies and even predicting the precursors to anomalies. All of these models require large investments in compute power as well as research time and effort. Certain aspects of model building and training are identical across any target application. The purpose of this project is to mature the capabilities of DAAS that help researchers perform critical yet time consuming tasks like formatting training data, scheduling jobs, and tracking model progress through multiple iterations.

Benefits: Our proposed tool can be used for a wide range of applications for NASA and other parts of US government where machine learning is being applied such as remote sensing application, weather data analysis and safety trends prediction. Researchers can save tine in building, testing and deploying these models and instead spend it on improving the models. The tool will also have the capability to share datasets and models.

The DAAS tool can help aviation consultants, industry analysts, and other firms investigating one-off or multiple machine learning problems. These users may analyze the impacts of new models by testing them against older models and assess their impacts before they are deployed.