I use it for managing high-throughput sequencing data. It's a good fit because a) the items are inherently key-value oriented and b) items have heterogeneous schema that are dependent on the platform used to perform the sequencing and the experimental origin of the samples.