In recent years, discovering novel drug-like molecules become a common application of machine learning. At the same time, it is much harder to discover molecules with properties optimized for being e.g. promising drug candidates. During the talk, I will present one of the main approaches to overcome this challenge – optimization with generative modeling.