The development of chip manufacturing and advanced packaging technologies has significantly changed redistribution layers (RDLs), leading to shrinking line width/spacing, increasing the number of build-up layers and package size, and introducing organic materials such as polyimide (PI) for dielectric. The fineness and complexity of structures, combined with the temperature-dependent and viscoplastic properties of organic materials, make it increasingly difficult to predict the thermo-mechanical behavior of wafer-level Cu-PI RDL structures, posing a severe challenge in warpage prediction. This study models and simulates the thermo-mechanical response during the manufacturing process of Cu-PI RDL at the wafer level. A cross-scale wafer-level equivalent model was constructed using a two-level partitioning method, while PI material properties were extracted via inverse fitting based on thermal warpage measurements. Warpage prediction results were compared against experimental data using maximum warpage as the indicator to validate the extracted PI properties, yielding errors under less than 10% at typical process temperatures. The contribution of RDL build-up, wafer backgrinding, chemical mechanical polishing (CMP), and TSV/TGV interposers to the warpage was also analyzed through simulation, providing insight for process risk evaluation. Finally, an artificial neural network was developed to correlate copper ratios of four RDL layers with wafer warpages for a specific process scenario, offering a potential direction for layout design optimization.