Current analyses of the Lyman-alpha forest assume the primordial
power spectrum of density perturbations follows a simple power law
form. We have used a large suite of numerical simulations to
calibrate a minimally parametric framework for the power spectrum,
dropping this assumption. I shall discuss our methods, including our
use of the statistical technique of cross-validation to avoid over-fitting the data. Finally, I shall present our results, using current Lyman-alpha data from SDSS, and our forecasts for BOSS data.