Like proteins, RNAs adopt diverse and intricate three-dimensional structures in order to function. However, high-resolution experimental RNA structures are difficult to obtain due the intrinsic flexibility of RNA molecules. While ab initio RNA structure prediction is computationally expensive, homology modeling can be adopted in cases where there is substantial sequence similarity to an existing RNA structure. Nonetheless, the significant drawback is that current RNA homology modeling approaches are either not general enough1-3 or are not designed to handle large insertions and deletions 4. To circumvent this, we present an alternative homology modelling approach that leverages on the hierarchial folding of RNA. Namely, we build on our previously developed Hierarchical Natural Move Monte Carlo5 protocol, and pull (through harmonic springs) the target sequence to the template structure by controlling the hierarchical degrees of freedom of the RNA molecule. In this way, we are able to make large conformational changes while preserving the physical geometry of any insertions and deletions, and retaining atomistic representation of all atoms. Additionally, because we control the molecular degrees of freedom based on RNA secondary structure, our protocol is general and easily scalable to larger RNA systems. We benchmark the performance of our approach to other existing techniques and also illustrate how we are able to successfully handle large insertions using SAM riboswitch as an example. By preserving atomistic representation of the RNA, our homology models can be directly used in RNA-based computational small-molecule drug design.