In N=2 Quantum Field Theories and Their BPS Quivers by Alim, Cecotti, Córdova, Espahbodi, Rastogi and Vafa the authors give a recipe which constructs from a pair $(C,\phi)$ consisting of a Riemann surface $C$ and a meromorphic quadratic differential $\phi$ on $C$ a (non-compact) Calabi-Yau 3-fold $X$.
Essentially, $X$ is cut out from the total space of the bundle $E=T^*C\oplus T^*C\oplus T^*C$ by the equation $uv=y^2-\phi(z)$ where $z$ is a local coordinate on $C$ and $(u,v,y)$ are coordinates on the fibre $E_z$. An important feature is that $X$ contains the spectral curve of $(C,\phi)$, i.e., the curve $\Sigma\subset T^*C$ which is a branched double cover of $C$ branched at the critical points of $\phi$.
From poking around physics literature that I don't understand, I get the sense that this construction is in some sense post hoc and the natural construction is to go the other way. That is, one should start with (some particular type of) CY $X$ and extract from it the pair $(C,\phi)$. Hence my question:
Is there a natural recipe which takes a CY $X$ and returns the pair $(C,\phi)$, inverse to the construction above?
Any insights or references to the literature would be greatly appreciated.