There are environments where you won't be sat at a desk all day. Many research or engineering environments will have you working both at your desk and in the laboratory or on the engineering floor.
For instance where I work I have an ordinary desk where I do admin. and intensive or off-line programming tasks, but I also have two experimental stations that I support. Whenever I need to test code on 'the machine' I have the option of using remote desktop or going down and working in the experimental 'hutch'.
In my previous three jobs I similarly had a desk, but much of the work was done down in the workshop or on the shop floor, testing, optimising or coding on the machines we were developing. I've even been sent out to the US, Germany, Taiwan and China to support and develop software further on machines in the field after they have been delivered to customers.
Even in my first job after university, which was more desk-bound than anything since, I was sent out to customer sites to deploy software and help with troubleshooting.
Incidentally, while much of my career could be described as embedded systems, as Thomas Owens suggests, don't assume that embedded means micro-controllers, in-house designed circuit boards or hard core operating systems like QNX or OS-9.
Much of industry runs on PLCs as Wolfger suggests and industrial PCs running operating systems like Linux, Windows Embedded or even just plain old Windows. These systems integrate 'hard real-time' subsystems with 'soft real-time' control systems. Where the top-most layer can be the most complex and challenging, especially when trying to represent a complex machine with a simple, easy to use user-interface.