Amazon DynamoDB is absolutely the easiest and most efficient way to run a Non-Relational database. It's perfectly suited for a "source of truth" where other indexes might be built from data stored in DynamoDB, and for use where search patterns are predictable. It's also incredibly effective at storing relatively unstructured data, or data models that might frequently change. As it does not require a rigid definition of data beforehand, it makes adjusting data stored in the table a non-effort, but it does require some planning when looking at how to get data back out of the table.
Unlike traditional SQL or Relational databases, DynamoDB is designed to have all relevant information within a single object. If not properly planned, this can lead to issues when building out a front-end. DynamoDB does not have join table support, nor does it support complex searches or "count" style responses. It is eventually consistent, although they recently did introduce Transaction support, there is no "rollback" option.