We consider the Langevin diffusion process of the form $dXt=−\nabla U(X_t)dt+\sqrt2\,dBt$, where $\nabla U$ is not necessarily Lipschitz continuous. We propose a discretization scheme that converges at the specified rate in terms of the 2-Wasserstein distance. Additionally, we demonstrate the application of the proposed method to the problem of sampling from log- concave distributions, a common task in machine learning. This presentation is based on joint work with M. Benko, I. Chlebicka, and J. Endal. For further details, see our papers: 2412.09698 and 2405.18034.