Modern approximate constructions of Ricci-flat metrics on compact Calabi-Yau manifolds typically fall into two complementary categories: highly flexible numerical methods, which are not easily interpretable, and analytic ansätze, which are often tied to a fixed Kähler class. This becomes particularly restrictive when seeking explicit Kähler moduli dependence, since varying the moduli generally requires changing the ansatz itself. I will describe a hybrid strategy that uses machine-learned Ricci-flat metrics as data and reconstructs explicit symbolic expressions with moduli-dependent coefficients. The resulting approximate analytic metrics exhibit explicit dependence on both Kähler and complex-structure moduli, providing a bridge between numerical metric learning and analytic approaches to Calabi-Yau geometry.