This study presents an innovative approach for estimating the proximate composition of diverse rice varieties using attenuated total reflectance fourier transform infrared (ATR-FTIR) spectroscopy and chemometric techniques. Principal component analysis (PCA) reveals distinct separations among the seven rice varieties based on their FTIR spectra. Robust partial least squares (PLS) regression models, developed with high calibration (R 2) values from 0.778 for protein up to 0.941 for moisture, demonstrate high accuracy in predicting proximate composition. The root mean squared error (RMSE) in percentage values, indicative of prediction accuracy, were low across all proximate components. To ensure the response variable of regression, proximate composition measurements were taken five times, while FTIR spectra were scanned tens of times, employing random numbers around the average with the same standard deviation as the measurement. Notably, the study emphasizes the pivotal role of the amide-III band in protein determination, alongside specific wavenumber regions associated with molecular changes in proximate components. This research underscores the potential of ATR-FTIR spectroscopy and chemometrics for rapid and accurate proximate assessment in food science and agriculture.