This dataset provides model code, input data, sample results, and documentation for an artificial intelligence-driven model, GeoCryoAI. GeoCryoAI is a hybridized process-constrained ensemble learning framework consisting of stacked convolutionally layered long short-term memory-encoded recurrent neural networks. The purpose of GeoCryoAI is to quantify permafrost thaw dynamics and greenhouse gas emissions in Alaska. The dataset includes pre-processed input data (i.e., thaw depth, active layer thickness, thaw subsidence; CO2 flux, CH4 flux) acquired from in situ measurements (e.g., CALM, GTNP, ITEX, SMALT STDM, ReSALT, AmeriFlux, NEON), remote sensing platforms (e.g., UAVSAR, AVIRIS-NG), and process-based modeling products. Field data were included to quantify CO2 and CH4 flux (e.g., chamber, eddy-covariance, and tall-tower measurements via flux tower networks) and active layer thickness (e.g., mechanical probing, borehole temperatures, ground-penetrating radar). These measurements were resampled to a 1-km grid, standardized, transformed, and assimilated into GeoCryoAI, a framework that simultaneously ingests, scales, and analyzes input data after resolving disparate spatiotemporal sampling and data densities. Model outputs were generated from two process-based models: SIBBORK-TTE derived thaw subsidence and TCFM-Arctic generated carbon flux outputs. The objective was to quantify how the Arctic is changing in response to climate change and how evidence of the permafrost carbon feedback may contribute toward a better understanding of the uncertainty of nonlinear feedbacks and their impact on the earth system.