Natural Killer (NKs) cells are a promising allogeneic immunotherapy for several malignancies. Their natural anti-tumor activity is orchestrated by a diverse array of activating and inhibitory receptors, enabling potent effector function against heterogeneous tumors while maintaining a favorable safety profile. However, sourcing NK cells from peripheral blood or umbilical cord blood units (CBUs) poses challenges due to unpredictable anti-tumor efficacy resulting from donor variability and limited cell quantity. We developed an integrated screening and manufacturing platform that leverages a proprietary small molecule, IBR403, to achieve 1E3-1E6 fold expansion of CBU-derived hematopoietic stem cells (xHSCs). This platform enables use of CBU segments, containing less than 300 µL of blood, while leaving the 25 mL CBU cryopreserved and intact for later use. We screened NK cells differentiated from expanded HSCs (xHSC-NKs, n>60 donors), then characterized their phenotype and anti-tumor function against 10 different tumor targets. Our machine learning (ML) models compute donor ranking based on genotypic, phenotypic, and functional characteristics from baseline assays. By acquiring hundreds of cytotoxicity measurements per sample, we identified associations between genotype, phenotype and function. Our data allowed us to train a penalized regression model using hundreds of features that predicts NK cell anti-tumor capacity. The model showed high accuracy at predicting donor cytotoxicity using parameters of HSC and NK phenotype and expansion. Our innovative expansion and screening method enables the prediction of a donor’s clinical performance using data obtained from a CBU segment. Importantly, this approach preserves the entire CBU for large-scale production, facilitating selective NK manufacturing. Furthermore, our lab-in-a-loop approach enables the in-depth study of NK cell functions through integrating tumor ligandome and NK cell receptor expression data, providing valuable guidance for designing clinical trials and matching patients to optimal donors.