Multi skill resource-constrained project scheduling Problems (MS-RCPSP) have been object of studies from many years. Also, preemption is an important feature of real-life scheduling models. However, very little research has been investigated concerning MS-RCPSPs including preemption, and even less research moving out from academic benchmarks to real problem solving. In this paper we present a solution to those problems based on a hybrid method derived from large neighborhood search incorporating constraint programming components tailored to deal with complex scheduling constraints. We also present a constraint programming model adapted to preemption. The methods are implemented in a new open source python library allowing to easily reuse existing modeling languages and solvers. We evaluate the methods on an industrial case study from aircraft manufacturing including additional complicating constraints such as generalized precedence relations, resource calendars and partial preemption on which the standard CP Optimizer solver, even with the preemption-specific model, is unable to provide solutions in reasonable times. The large neighborhood search method is also able to find new best solutions on standard multi-skill project scheduling instances, performing better than a reference method from the literature.