N-terminal coding sequences (NCS) are key regulatory elements for fine-tuning gene expression during translation initiation, the rate-limiting step of translation. However, due to complex combinatory effects of NCS biophysical factors and endogenous regulation, designing NCS remains challenging. Herein, we implemented multi-view learning strategy for model-driven generation of synthetic NCS for Saccharomyces cerevisiae and Bacillus subtilis, which are model microorganisms widely used in the laboratory and industry. Overall design: NCS expression level was measured using a “FACS-seq” reporter-based assay. Libraries of yeast and bacteria cells harboring a plasmid in which GFP expression was driven by Pbs, and randomly-generated NCS were inserted after the start codon (ATG) of GFP. Cells of each strain were FACS-sorted into high and low bins on the basis of expression. Specifically, NCS libraries for S. cerevisiae and B. subtilis has nearly 150,000 cells that were sorted, respectively.