Kinetic equations usually arise from mean-field limit of N-body particle systems. As such they contain, inevitably, uncertainties in their collisional kernels, scattering coefficients, initial and boundary conditions, and forcing and course terms. Conducting uncertainty quantification for kinetic problems is important to assess and calibrate the kinetic models. We will develop efficient computational methods, based on multi-level Monte-Carlo methods and multi-fidelity approaches, and also study Bayesian inference based inverse problem in order to determine, in probabilistic framework, uncertain coefficients and data in kinetic models.