MLSysOps work is structured in 6 work packages as follows:
WP1 is responsible for the administrative and technical management of the project.
WP2 focuses on the requirements analysis and the design of the MLSysOps framework.
WP3 is responsible for the development of the AI-ready framework and its underlying mechanisms, while WP4 is responsible for developing the ML architecture and ML models that will drive these mechanisms. Both the framework and the ML models are evaluated in WP5, which also drives data collection. Finally, WP6 is responsible for all activities that contribute to the project’s impact.
The chronological lifetime and main dependencies between the work packages are shown in the figure below. The results of WP2 feed into WP3 and WP4 to guide the development of the MLSysOps framework and ML architecture and models, as well as into WP5 to guide the evaluation of the WP3 and WP4 results. There is also a dependency between WP3 and WP4 since the ML models will be integrated with the AI-ready mechanisms, as well as between WP5 and WP4 since the former provides data that will be used to train the ML models developed in the latter.