The Mw 5.5 Karakanja earthquake, which occurred on 24 January 2020 in Tajikistan, is a significant moderate seismic event within the tectonically active Pamir region. This area is characterized by the complex oblique convergence between the Indian and Eurasian plates, leading to intricate deformation patterns. This study comprehensively investigated the coseismic deformation field and detailed earthquake source parameters of the Karakanja earthquake using high-resolution Sentinel-1 Interferometric Synthetic Aperture Radar (InSAR) data and a sophisticated Bayesian inversion methodology. Interferograms from ascending and descending satellite tracks revealed surface displacements along the satellite’s line-of-sight (LOS), with maximum values of approximately 4.7 cm (ascending) and 6.7 cm (descending). These measurements highlight substantial ground deformation concentrated around the causative fault. Applying the Bayesian inversion technique to these InSAR data allowed us to define the geometry and slip distribution of the causative fault. The inversion results indicate a preferred shallow fault plane with a modeled fault area of approximately 23 km². Key parameters include a centroid depth of 6.2 km, a strike of 29°, and a dip of 32.7°. Analysis of slip components revealed a dominant dip-slip component of 0.46 m and a strike-slip component of 0.15 m. This suggests a predominantly reverse faulting mechanism with a minor right-lateral component (rake angle ~72°), aligning well with the regional tectonic regime of oblique convergence. To rigorously assess model quality, statistical and spatial analyses of residuals were performed. The inversion yielded a root mean square error (RMSE) of 5 mm and an R² value of 0.92, indicating the model explains approximately 92% of the observed deformation. Residual analysis confirmed the absence of systematic bias, with high correlation (ρ=0.96) between observed and modeled displacements. Spatial autocorrelation analysis (Moran’s I = 0.14, p < 0.001) revealed weak but significant spatial autocorrelation in residuals, suggesting that while the uniform-slip model captures the main deformation pattern, it acknowledges minor small-scale complexities or residual atmospheric effects. Geostatistical analysis showed a spatial structure in residuals near the fault, indicating that the uniform-slip model inadequately represents local complexities. The estimated seismic moment was 2.25 × 10¹⁷ N·m, with a stress drop of ~5 MPa. These values align with empirical scaling relationships for continental crustal earthquakes. The modeled fault geometry and slip mechanism are fully consistent with the regional oblique convergence tectonics of the Pamir region. This study demonstrates the effectiveness of Bayesian inversion for determining source parameters of moderate earthquakes, while employing a multi-level validation framework (including spatial residual analysis and comparison with empirical relationships) to assess uncertainty and the "preferred" nature of the fault plane.