<p>The Northeast Himalayan region of India is one of the most seismically active and tectonically complex area globally, but the scarcity of recorded strong ground motions poses challenges for reliable seismic hazard assessments. This study introduces a refined site correction methodology based on the Empirical Transfer Function (ETF) to improve the accuracy of synthetic ground motion simulations in data-deficient environments. The ETF was estimated for 48 sites across the Northeast Himalaya using 447 accelerograms from 28 earthquakes (Mw 3.9–5.9). Incorporating the ETF into the stochastic simulation process significantly improved the agreement between synthetic and recorded time histories, particularly in the high-frequency range (0.2 to 10 Hz), compared to conventional horizontal-to-vertical ratio-based corrections. The spatial distribution of dominant frequencies derived from the ETF provides new insights into local amplification characteristics. Quantitative error metrics validate the approach, with time-domain and frequency-domain errors reduced from 1.72 to 0.36 and 0.612 to 0.092, respectively. The ETF-based methodology enhances the realism of synthetic accelerograms, strengthens regional seismic hazard evaluations, and contributes to resilient infrastructure planning in the Northeast Himalaya. This scalable framework can be extended to other tectonically active regions with similar geological complexities.</p>