Hidden hunger, also known as micronutrient deficiencies in pregnancy, is a very dangerous risk to both the mother and the child. Current prenatal nutrition models are based on generalized nutritional guidelines and cannot be adjusted to genetic, physiological and environmental differences. IoT data, dietary records, genetic research, and medical imaging data are still disjointed, which constrains the number of personalized care opportunities. In order to close these gaps, we present a real-time AI-based prenatal nutrition system, which combines multimodal data, such as wearable IoT streams, genetic and epigenetic profiles, dietary records, and 3D/4D ultrasound images. The system uses an ensemble deep learning system that integrates U-Net, ACSNet, EU-Net, MSNet, and PEFNet to give adaptive dietary suggestions, and CapsNet facilitates dynamic fetal health. The constant IoT feed guarantees a dynamism in real-time, and ultrasound imaging improves anatomical examination. Experimental outcomes show that the 94% accuracy of predicting the micronutrient deficiencies significantly outperform the traditional use of single-model methods, with the use of ultrasounds enhancing the accuracy of fetal monitoring by 31 percent. This framework provides early personalized nutrition interventions timely, to promote prenatal care using the precision health strategies to maximize the maternal and fetal outcomes.