ABSTRACT
Data infrastructure as a term has wide-ranging interpretations across communities. The authors can embed analysis, intelligent decision-making, and instrument control into one integrated AI/ML module. When that module hits a bug, it will have to be fixed and the entire pipeline has to be restarted. Second, control, analysis and decision-making can all run separately and only be connected by a simple communications protocol. When the input is received, the module performs its designated task and the result is sent to the next module. When one of the modules hits a bug, the other components keep running, the bug can be fixed and only the affected module requires a restart. If a module has to be switched out during an experiment or in between experiments, the others will be unaffected. The only change is that the instrument has to listen to messages coming from the decision-making module and send messages where the data analysis is executed.
