Moisture content is one of the factors measured to evaluate the quality of Camellia oleifera seeds. High quality <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">C. oleifera</i> seeds used for trading must have a low moisture content, specifically not more than 15% on a dry basis (db). Moisture content analysis requires a prolonged laboratory investigation so that the development of fast and effective determination methods is helpful. The objective of this paper was to develop a low-cost portable NIR reflectance spectrometer collaborating with an android application for the rapid prediction of the moisture content in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">C. oleifera</i> seeds. To calibrate the prediction model, an effective chemometric algorithm, based on partial least squares regression was established, and models based on wavelength selection algorithms such as backward interval partial least squares ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">bi</i> PLS) and partial least squares coupled with variable importance projection (VIP-PLS) were implemented as an improved version of PLS. Both algorithms ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">bi</i> PLS and VIP-PLS) improved the predictive performance and accuracy of the model. The experimental results showed that the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">bi</i> PLS model with the 1 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">st</sup> derivative transformation provided the best prediction for measuring the moisture content of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">C. oleifera</i> seeds with a coefficient of determination (R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) value of 0.927, standard error of prediction (SEP) of 0.848%db, bias of −0.067%db, function slope of 1.005, and ratio of performance deviation (RPD) of 3.696. Finally, the device was tested according to the ISO 12099:2017(E) standard and confirmed the reliability of the device for in-field use.