Quantlib has C# bindings so you should be able to use it. If you are having problems getting it to work, you might want to look at the work done by Resolver Systems (the maker of an IronPython/.Net-based spreadsheet product) to integrate Quantlib with their .Net API here.
Alternatively, consider JQuantlib a port of Quantlib's C/C++ code to Java, which you could then pass through IKVM to get .Net assemblies (although I wouldn't want to comment on how .Net idiomatic the resulting API would be).
BUT as great as having an open-source quant library is, I don't know to what extent it is being used to make 'real money'. You may want to seriously consider the commercial alternatives (for which source code may not be available); Numerix is probably one of the most popular but there are many others.
Also, production quality versions of these libraries almost always tend to be written in C++. Without wanting to get into performance arguments, 'quants' generally believe that JIT-ed languages (Java, .Net) are inefficient (both in terms of performance and expressive capability) for describing the complex maths behind today's financial product pricing (again, not wanting to get into an argument about the social virtue of today's derivative products!). That said, I believe Credit Suisse has written significant chunks of their analytics in F#.
I am a little confused as to what your real requirements are here; you state that you are at a large bank, are looking for a reasonably comprehensive quant library... in my experience your organisation would already have built/bought one?