How to efficiently aggregate data in billions of individual records in AWS?

Question

At a high / theoretical level I know exactly the type of architecture I want to build and how it would work, but I\'m attempting to construct this as cheaply as possible usi