In CEST/NOE experiments, induced saturation intended for certain metabolites also saturates other moieties, leading to mingled backgrounds in Z-spectra. Precise quantification of endogenous metabolites is impeded by slow acquisition times and confounding background contributions. We present an optimization framework targeting a flattened baseline in Z-spectra, with direct water saturation and non-specific macromolecular contributions fully compensated. Saturation pulse amplitudes as a function of frequency offset were derived using optimal-control-based calculation. Feasibility of this approach is demonstrated in vitro for NOE/CEST in the presence of tissue-like background and accelerated acquisition by distinct offset selection. Experimental results show good agreement with concentration levels.