Lumped thermal airflow models, widely used in whole building simulation tools, are inadequate for capturing thermal stratification that typically arises while using passive, low-energy, heating and cooling systems (e.g., displacement ventilation). Computational fluid dynamics (CFD) on other hand is intractable for practical design and optimization, due to the large dimensions of the state-space (often on the order of a few millions) obtained on numerical discretization of the governing fluid equations. Even though the dimension is large, the behavior of the flows itself is often low-dimensional. Data-driven model reduction methods are important to obtain low-order models that accurately describe the original dynamics and are tractable for control design. In this talk, I will present some algorithms for model reduction, based on a control theoretic method called balanced truncation. The method will be illustrated with an application to a room equipped with displacement ventilation. We develop models for capturing relevant airflow dynamics, and develop controllers that maintain occupant comfort while minimizing energy consumption. If time permits, I will also present an application to the control of an airfoil wake at low Reynolds numbers, with the motivation being the design of micro-air vehicles that are robust to sudden gusts in flight.