For years, AI in drug discovery has been held back by a deceptively simple problem: the data isn’t good enough. Mountains of sequencing, pooled perturbation studies, and those mixed-cell experiments ...