I've been working on a suite of modules for the past couple of months to make building applications on top of Amazon's SimpleDB as easy as building applications on DBIx::Class. It's called SimpleDB::Class.
It gives you an object relational mapper style interface to SimpleDB, which is heavily borrowed from DBIx::Class. It uses Memcached to speed up queries, increase capacity, and get around most of the stickiest problems with eventual consistency.
SimpleDB::Class also hides all of the things that most new developers find difficult about SimpleDB, things like cascading retries, next tokens, and the pseudo rest-like protocol for interfacing with SimpleDB. Not to mention handling searchable/sortable dates and numbers. As a developer, you simply give it some data or ask it for some data, and all those things just happen automatically behind the scenes.
There are even many nice utility functions like max and min, which are posted in questions to the SimpleDB forums all too often.
I've still got a decent todo list for it (things like cnames, pagination, and auto-sharding), but it's a good working solution right now, that's being used in our environment.
If you have a wishlist for features, or better yet you have patches, let me know. I'm happy to include them. You can download it from the CPAN, or read about it online at: http://search.cpan.org/~rizen/SimpleDB-Class/lib/SimpleDB/Class.pm