My thesis addresses a fundamental gap in sleep wearables: devices collect extensive physiological data but fail to translate it into meaningful insights that drive actual behavioral change. I'm building an interpretable machine learning system that treats sleep as a multidimensional cognitive process involving homeostatic pressure, circadian timing, autonomic balance, and memory consolidation. By anchoring algorithmic outputs in established neurocognitive theory, my system provides users with mechanistic explanations of their sleep patterns rather than opaque scores. This approach frames raw wearable data as evidence for testable hypotheses about underlying brain processes, helping users recognize cues, understand routines, and identify rewards—the key elements for habit modification that current wearables miss. The result makes cognitive mechanisms visible, giving people clear, actionable feedback about what influences their sleep and concrete steps to improve it.