In my talk I will discuss two applications of machine learning that could be used for investigating the nature of Dark Matter (DM) through modern astronomic observations. Firstly, I will focus on the possibility of detecting dark galactic subhalos (i.e. DM subhalos that are essentially devoid of baryons) through their gravitational influence on the Milky Way's stellar field. Their successful detection would provide strong evidence in favor of the cold DM paradigm while the opposite is true in case of their absence. In the second part of my talk I will focus on the problem of inferring the DM content of local galaxies. This has been traditionally done through the use of galactic rotation curves, however, such approach does not utilize all of the information contained in observations. I will present an alternative approach in which raw observational images could be directly analyzed by a convolutional neural network that has been trained on mock observations generated from hydrodynamic simulations.