Evolutionary test generation techniques have emerged as a popular approach in recent years for enhancing the testing of software systems. However, while these techniques have proven to be efficient in programming languages that support static type annotations, dynamically typed programming languages have not received significant attention from the automatic test generation community. This paper introduces an approach aimed at automatically generating fully executable unit tests suitable for dynamically typed programming languages. In particular, our approach is tuned for dynamically-typed and class-based programming languages, and it is implemented in the Pharo and GToolkit programming languages. To address the absence of static type annotations, our approach uses a type profiling mechanism and employs a genetic algorithm to drive the evolution of the unit tests.