RR Lyrae stars are one of the best tracers of the oldest populations of stellar systems. The RR Lyrae content of the low-latitude regions of the Galactic disk is especially interesting in light of the formation of the thick disk. Unfortunately, severe extinction has prohibited the exploration of the variable star content of this region so far, but with the completion of the Vista Variables in the Via Lactea (VVV) survey, we can bring light to these questions. Machine-learning techniques have been adopted to classify RR Lyrae variables based on their light-curve shapes in the light curves of the approximately 200 million point sources in the VVV, resulting in the detection of 1892 high-confidence RR Lyrae stars. Utilizing newly developed techniques, we have measured mean magnitudes of these variables, enabling the determination of accurate distances for each of them. Furthermore, we show that the mean iron abundance of RR Lyrae variables can be determined from their K-band light-curve shapes, which we use to explore different sub-populations of the newly found variables. The structure of the metallicity distribution of this sample bears remarkable similarity to that of the ARGOS survey, covering the much younger red clump stars of the Galactic plane and bulge. The few differences can be easily understood as a lack of RR Lyrae stars with supersolar metallicities, as well as halo interlopers enhancing the RR Lyrae metallicity distribution at the lower metallicity end. The weight of each sub-population evolves with Galactocentric distance, supporting inside-out formation scenarios.