Qilin watermelon is a unique and sought-after variety known for its rich flavor and high market value. Motivated by SDG 2 (Zero Hunger), this study aimed to improve the efficiency and precision of ripeness classification to reduce post-harvest losses and improve food security. Using deep learning and machine learning techniques, this work explored a two-phase approach using visual data and tapping sounds of watermelons. In the image classification phase, the VGG-16 model is fine-tuned with data enhancement, achieving 85% test precision. For audio-based classification, spectral imaging yielded an accuracy of 84%, while the extraction of MFCC features combined with a random forest classifier achieved an exceptional accuracy of 98%. The novelty lies in the integration of diverse data modalities for ripeness prediction, enhancing reliability. These promising results underscore the potential for scalable applications in agriculture, helping stakeholders optimize harvest and reduce waste. However, challenges remain with the generalizability to other watermelon varieties and its sensitivity to noisy audio data. Future efforts will address these limitations by expanding dataset, improving robustness, and optimizing the model for efficient hardware deployment.