Twitter Bootstrap; Knockout; Rails 3; Postgresql for 'standard' models; Redis for training-stats + background tasks; Capybara + rspec + semaphore (CI); git (github); Linode; Vim / Textmate; ...and IBM Model-M keyboard :)
P.S. I hope that you have your own closed office for that clicky keyboard :)
Redis is fantastic. Perhaps I should also mention we use ohm as an ORM, which is very neat. I wouldn't dream managing our particular app without it. But it's not exactly fire-and-forget. Firstly, modelling your data is very important, or you can end up with lots of stuff taking precious memory. Secondly, whilst performance is great, it's still not 'transparent' like accessing a variable in memory. If you have lots of small redis queries, they do add up eventually. This is getting very detailed, but you might want to check out this as an example (specific to the ruby version it seems): https://github.com/redis/redis-rb/issues/283
Hope this helps, but feel free to get in contact.
We used bootstrap as a starting point and - to be honest - at that point it looked like this. But over time we changed and added more and more of our own code. But I would say it is quite marginal still if you look at what bootstrap provides.
I like and we still use twitter bootstrap for the grid, the readability of the text and the basic layout elements. But it is very important to give it your own style and diverge from it over time.
Of course the anatomy trainer itself is mainly built without bootstrap. But even here we use elements of it.