Mission Relevant ML Artifact Registry

Data driven AI/ML approaches are emerging as a dominant method for addressing difficult problems, and both academia and industry are providing state-of-the-art solutions to many problems in technical and scientific domains. These communities make many of these reference implementations accessible, leading to opportunities for the overarching community to leverage existing work and spur innovation. A similar approach within the DoD community will provide similar benefits to mission relevant problems. For this proposed effort, Black River will create interfaceable repositories containing state-of-the-art algorithm implementations using the Streamlined ML Model Integration Software Toolkit. These interfaces will provide role-based opportunities for developers, curators, and operators for accelerated research, development, and application of data driven approaches. We will deploy into a cloud environment, conduct a survey of existing artifacts and use cases, develop reference implementations, and perhaps most importantly develop and demonstrate an enhanced user interface. This piece is critical, as most existing capabilities are geared towards developers, while ultimate mission utility comes with engaging operators. The core innovation on this effort will be the translation of role-based requirements into the interface to support the defined use cases. This will include enhancing the developer interface and creating both a curator and operator interface.