The DEEP-HybridDataCloud project researches on intensive computing techniques such as deep learning, that require specialized GPU hardware to explore very large datasets, through a hybrid-cloud approach that enables the access to such resources. DEEP is built on User-centric policy, i.e. we understand the needs of our user communities and help them to combine their services in a way that encapsulates technical details the end user does not have to deal with. DEEP takes care to support users of different levels of experience by providing different integration paths. We show our current solutions to the problem, which among others include the Open Catalog for deep learning applications, DEEP-as-a-Service API for providing web access to machine learning models, CI/CD pipeline for user applications, Testbed resources. We also present our use-cases tackling various problems by means of deep learning and serving to demonstrate usefulness and scalability of our approach.

DEEP HybridDataCloud receives funding from the European Union's Horizon 2020 research and innovation programme under agreement RIA 777435.

Trabajo presentado al ISC High Performance, celebrado en Frankfurt (Alemania) del 16 al 20 de junio de 2019.