If you're not buying into vaporware, this has been existing for a while and proven on large projects. It supports the world's best optimization engines, it's embedded in Java and supported under Eclipse, and comes with extensive documentation and samples library:
http://www.ateji.com/optimj
What is important in having an embedded declarative modeling language is that it requires very little learning (we have seen Java developers running their first optimization model in 15 minutes), that it removes the pain and the bugs of translating data formats back and forth, and that you benefit from the existing libraries and productivity tools from the underlying platform.
In the case of OptimJ, you would typically use, directly from your optimization model, JDBC or Hibernate for database access, JExcelAPI for reading/writing spreadsheets, SWT or Swing for graphical interfaces, JavaDoc for automated documentation, JUnit for regression testing, and so on. The same obviously applies for MSF and the .NET platform.
Additionally, OptimJ provides concise and powerful bulk data manipulation primitives, because typically 80% of model code happens to deal with input data formatting and cleaning.