I've seen people discuss on HN how they've used Amazon MechTurk to do this as well. I believe their conclusion was that they weren't particularly happy with it.
Realistically, if you're looking for tech compatibility, you may need to do it yourself. It's not as bad as it seems, but you do have to test 3-6 browsers/OS combos. If you've used good stuff like jQuery, etc., lots of the browser differences can be made minimal.
Also, you're probably not going to get particularly phenomenal feedback from just any user. Try doing the testing in person, with just a few people (Jakob Nielsen has a post about how many people you need, and concludes something like 3 people will give you the majority of what you need to focus on. After all, you want to focus on the big issues, and as a startup you're likely to have tight resources available...)