My research programme is guided by the widely resisted idea that the human brain is too complex to be rendered either intelligible or actionable by the parsimonious, low-dimensional models dominant in the field. If clinical translation is the ultimate objective, this position compels a two-pronged approach at either extremes of data-scale: single-subject studies employing individually adaptive models and interventions, and large-scale studies employing high-dimensional inference powered by machine learning. It also compels a methodological flexibility, targeting clinical, scientific, or operational objectives as the fastest route to impact dictates.