In this study, we proposed and evaluated a digital micro-mirror (DMD) based ELICO near-infrared spectrometer (NIRS) for predicting moisture, protein, and fat content in soya meal for poultry feed at a low cost and compared the results to those of the Bruker NIRS. Preprocessing, partial least squares regression (PLSR) methods, and the wet chemical method were used to develop prediction models on both instruments by using soya samples with moisture variations of 9–16%, protein variations of 41–51%, and fat variations of 0.7–3.0%. We found that the wavelength ranges of 1410–1470 nm, 1470–1560 nm, and 1100–1400 nm are sensitive wavelength ranges for the precise detection of moisture, protein, and fat, respectively. The experiment results of ELICO NIRS showed that the correlation coefficient (R2), root mean square error (RMSE), relative performance determinant (RPD), and range error ratio (RER) were 0.89, 0.77%, 3.2, and 8.5; 0.84, 1.03%, 2.7, and 9.7; and 0.86, 0.29%, 2.7, and 7.8 for moisture, protein, and fat, respectively. The maximum standard deviation was 0.006, 0.005, and 0.006 found for moisture, protein, and fat, respectively. Bruker NIRS showed that the R2, RMSE, RPD, and RER were 0.83, 1.03%, 2.4, and 6.4; 0.89,0.75%, 3.0, and 13.5; and 0.74, 0.31%, 2.7, and 7.6 for moisture, protein, and fat content of soya meal, respectively. Further, the prototype outcomes were compared with the previous studies done with various techniques. As a result, the ELICO NIRS has a good figure of merit that fits within an acceptable range, and the prediction findings were well correlated with the wet chemistry method, performed similarly to Bruker NIRS. The ELICO prototype can reliably estimate a wide range of feed ingredients and has a lot of potential as a field and laboratory instrument, allowing small-to-large industries, retailers, and educational institutions to use it at a low cost.