These findings present a proof-of-concept for using non-invasive closed-loop technology to optimize human learning and memory through principled stimulus timing, but only in those participants for whom classifiers reliably predict out-of-sample memory function.