Statistical-physics models constructed directly from data have succeeded in describing the collective phenomena in many biological systems. While much success has been achieved in learning the steady state distribution, going into the (especially out-of-equilibrium) dynamics is more challenging, for example with problems such as causality, co-existing of many time scales and non-ergodicity, asymmetric interaction, partial and finite information, and dynamically-evolving landscapes. In turn, the environmental drive can also be used as an active perturbation to elicit more information. This internship project will address this problem by starting in one example of inferring interactions for non-equilibrium dynamics, to be discussed according to the expertise and the interest of the applicant. Specifically, the internship will include inferring interaction models from real biological data from neuroscience and/or collective animal behavior, and investigating toy models of interacting spins.