Understanding hierarchical self-assembly of biological structures requires real time measurement of the self-assembly process over a broad range of length- and timescales. The success of high-speed atomic force microscopy (HS-AFM) in imaging small scale molecular interactions has fueled attempts to introduce this method as a routine technique for studying biological and artificial self-assembly processes. Current state of the art HS-AFM scanners achieve their high imaging speed by trading achievable field of view for bandwidth. This limits their suitability when studying larger biological structures. In ambient conditions, large range scanners with lower resonance frequencies offer a solution when combined with first principle model-based schemes. For imaging molecular self-assembly process in fluid however, such traditional control techniques are less suited. In liquid, the time-varying changes in the behavior of the complex system necessitates frequent update of the compensating controller. Recent developments in data-driven control theory offer a model-free, automatable approach to compensate the complex system behavior and its changes. Here we present a data-driven control design method to extend the imaging speed of a conventional AFM tube scanner by one order of magnitude. This enabled the recording of the self-assembly process of DNA tripods into a hexagonal lattice at multiple length scales.