Understanding regional variations in post-fire vegetation recovery is essential for enhancing forest resilience and informing land management under changing climatic conditions. This study combines multi-source satellite time series, including NDVI, EVI, SAVI, NBR, LST, ET, and drought indices, within a Google Earth Engine workflow and applies a Multi-Layer Perceptron neural network to model post-fire vegetation dynamics across twelve wildfire events in the United States and Spain between 2000 and 2022. The model achieved high predictive accuracy, with R² values exceeding 0.95 and RMSE values below 0.015 across all sites, including independent test fires, demonstrating its capacity to capture non-linear and region-specific recovery patterns. Two quantitative indicators—Vegetation Recovery Rate and Vegetation Recovery Period at 50, 70, and 90 percent of pre-fire NDVI—revealed pronounced heterogeneity among ecosystems. Some forested regions recovered within two to five years with VRR values up to 0.035, whereas Mediterranean and semi-arid areas displayed extended or negative recovery trends with VRR values as low as −0.009 and VRP90 reaching approximately 12 years. Results showed substantial variability in recovery trajectories driven by ecosystem type, climat,e and fire severity. While forested regions exhibited faster recovery, Mediterranean and semi-arid ecosystems experienced prolonged recovery or degradation trends. This framework provides a scalable and transferable approach for monitoring post-disturbance vegetation resilience using multi-source satellite time series and deep learning, with significant implications for post-fire restoration, adaptive management, and climate-informed decision making.