This study addresses the search for Carbon-Enhanced Metal-Poor (CEMP) stars based on low-resolution spectra and proposes a machine learning solution based on deep learning/neural networks. From the LAMOST DR8 spectral library, this work discovered 12,766 CEMP star candidates. For ease of reference and use, we provide the estimated parameters Teff, log g, [Fe/H], and [C/H] for the targets found in the search.