Estimation of the postmortem interval (PMI) remains a major challenge in forensic science. We used attenuated total reflection (ATR)–Fourier-transform infrared (FTIR) spectroscopy combined with chemometric modeling for PMI prediction using vitreous humor samples from 20 forensic cases with known PMI (24.8–97.6 h) and 10 with unknown PMI. The intensities of vibrational bands commonly associated with PMI were analyzed, and several peaks in the carbohydrate/phosphate region showed significant correlations with PMI. Principal component analysis revealed time-dependent spectral evolution, with PC1 (48.1%) associated mainly with carbohydrate/phosphate variations and PC2 (37.6%) with protein structural changes. Partial least squares regression with two latent variables achieved a cross-validated RMSE of 15.8 h (R² = 0.53) on all 20 known samples. Variable importance analysis identified glycoprotein degradation (1190 cm⁻¹) and phospholipid breakdown (736 cm⁻¹) as the dominant predictors, with traditional carbohydrate bands playing a secondary role. Predictions for unknown samples ranged from 27.1 to 80.1 h, with five of ten falling within the 90% prediction interval (±20 h) of the available estimates. This study presents a promising PMI estimation model that performed well on unseen samples, even if the sample size represents a methodological limitation that will be addressed in future investigations through larger, more diverse datasets.