Adaptive Just-in-Time (JiT) compilers employ a wide range of techniques to concentrate compilation efforts in the most promising spots of the application. The goal is to balance tight compilation budgets with appropriate level of code quality. Such techniques include profiling to identify hot spots in the application, using multiple compilation strategies, and performing aggressive optimizations. Some compiler researchers propose that JiT compilers should benefit from method-specific compilation strategies. These strategies can be discovered through machine-learning techniques, where a compilation strategy is tailored to a method based on the method's characteristics. This presentation describes our investigation of the use of Support Vector Machines to create a learning-enabled, method-specific compilation infrastructure in Testarossa, a state-of-the-art commercial JiT compiler employed in the IBM J9 Java Virtual Machine. This new infrastructure allows Testarossa to quickly explore a large number of different compilation strategies, and to generate the data needed for training such models. The infrastructure also enables Testarossa to integrate with a trained machine-learned model that predicts which learned compilation strategy has the appropriate balance of code quality and compilation effort on a per-method basis. We will present the results of an extensive experimental evaluation of the infrastructure and compare these results with the performance of the original Testarossa using adaptive operation, which was highly tuned by compiler experts over the course of many years.