Cytokine-induced memory-like (ML) NK cells have enhanced anti-tumor activity (e.g. cytotoxicity, IFNg) compared to conventional NK (cNK) cells. ML NK cells form after an initial brief activation with IL-12, IL-15, and IL-18 (IL-12/15/18) followed by a differentiation period to acquire ML properties. In a phase 1 clinical trial in leukemia, ML NK cells induced clinical responses in ~50% of patients compared to ~25% in historical clinical trials of cNK cells. Recent work revealed new heterogeneity of ML NK cell biology, where IL-12/15/18 activation and differentiation induces only a subset of cNK cells to become genuine ML NK cells, termed enriched memory-like (eML). eML NK cells produced the most IFNg following stimulation with leukemia cells, and conventional CD56^bright and CD56^dim NK cells gave rise to unique subsets of eML NK cells. Using scRNA-seq, we demonstrated that both eML subsets persist for at least two months within patients receiving ML NK therapy. Given the therapeutic potential of eMLs, as well as the specialized expertise necessary to identify NK cell subsets, we present an unbiased, reproducible, and user-friendly machine learning classifier for identification of eML and cNK cell subsets in droplet- based scRNAseq/CITE-seq data. This software, implemented in Python, builds upon our previously-published classifier. Our updated version is trained on an expanded set of “ground-truth” labeled scRNA-seq and CITE-seq datasets consisting of cNK cells, eMLs, short-term cytokine activated, and IL-12/15/18 activated and differentiated NK cells. Given its more diverse array of training data, our new classifier has the potential to make finer distinctions between NK subsets, and to distinguish between short-term activation and true memory-like states. This tool will greatly improve the identification of eML NK cells for therapeutic and research use, facilitating correlative studies on the phenotype and persistence of eML NK cells across clinical trials.