We present our recent findings (including published ones) on the real-time control of magnetohydrodynamic (MHD) flows—fluid flows driven by electromagnetic fields. Controlling fluid flows is inherently challenging due to their high-dimensional, nonlinear, and often chaotic nature. These complexities conflict with the constraints of traditional model-based control, which typically relies on embedded hardware with limited computational power and strict sampling rates. The addition of electromagnetic fields introduces another layer of complexity to the problem. Recently, fluid dynamics has benefited from the emergence of machine learning techniques for developing tractable models of fluid flows. One particularly promising approach is Koopman operator theory (KOT), which enables the representation of complex dynamics through a linear—albeit infinite-dimensional—operator, making its finite-dimensional approximations especially well-suited for control applications. In our recent work, we applied KOT to develop a control strategy for an MHD flow of a conductive liquid, actuated by a system of electrodes and coils, with feedback from Particle Image Velocimetry (PIV)—a technique that measures fluid velocity by tracking the movement of seeding particles in a camera feed. Our approach not only provides a systematic framework for controlling fluid flows by shaping external electric and magnetic fields but also represents the first experimental validation of data-driven control in fluid dynamics. Our results can enable a wide range of applications, from enhancing heat transfer to improving mixing efficiency.