<p>This paper proposes a gamma-ray-based multiphase flow monitoring system employing Monte Carlo simulation integrated with machine learning techniques. The system employs 12 CsI(Tl) detectors and 3 collimated ¹³⁷Cs gamma-ray sources for analyzing various multiphase flow regimes consisting of air, water, and oil. A deep learning model was developed using PyTorch to estimate the phase fractions of the flow regimes. The trained model achieved high accuracy (with an MSE below 1.2 on the test dataset and an <i>R</i>² score exceeding 0.91 over whole targets) in estimating phase fractions. This proposed approach demonstrates a powerful tool for real-time flow monitoring applications.</p>