A high-precision nanoscale characterization technology for multilayer films is critical in microelectronics, optoelectronics and bio-medicine. Ultrafast photoacoustic technology provides non-contact structural profiling with sub-nanometer resolution. In this study, an advanced scheme combining ultrafast photoacoustic method and Artificial Intelligence (AI) is applied to automatically measure the thicknesses and crystal orientations of multilayer thin films and superlattices. We developed a dataset using a multilayer photoacoustic theoretical model consistent with experimental results. To mitigate experimental noise, we applied a variational mode decomposition (VMD)-backpropagation neural network (BPNN) algorithm and an AlexNet framework for samples properties prediction. Characterization results in SiO2, LiNbO3 multilayers and GaAs/AlAs superlattices verify that this AI-based scheme can automatically get the knowledge of multiple properties with a higher precision in principle. This method enables tomographic detection of complex nanostructures and offers a novel approach for real-time monitoring of integrated devices and biomedical imaging.