I'm afraid there are some innacuracies
1. Aggregations can be stored within DynamoDB, you just need to setup how those are calculated, usually with DynamoDB streams and Lambda. You do not need to do it within the application at request time.
2. "Local Indexes" isn't entirely accurate. Its called a Partition key and you can't avoid using a partition key, its kind of the pivotal query mechanism of a table.
3. 10GB is not the limit of a partition. AWS's own documentation states that once a partition approaches 10GB it will be split.