India is highly vulnerable to natural disasters and the massive flow of multilingual and multimodal social media data during such events makes it difficult to extract timely and reliable information. To address this challenge, this paper introduces ALERTIFY, an AI-driven multimodal framework designed to convert unstructured social and news media content into actionable disaster intelligence. The system follows a three-stage pipeline: first, an Artificial Neural Network (ANN) combined with TF–IDF is used to filter disaster-relevant text; second, a Multinomial Naïve Bayes classifier categorizes events into disaster types such as flood, earthquake, cyclone, fire, and accident; and third, a YOLO-based object detection model analyzes video frames to detect people, fire, vehicles, and debris, while audio streams are processed for acoustic events and speech-to-text transcription. The framework was evaluated on custom datasets, with the ANN achieving an accuracy of 84.35% (F1-score 0.82) and the Naïve Bayes classifier reaching 88.89% accuracy. These results show that ALERTIFY can provide reliable situational awareness by integrating text, image, video, and audio into a single evidence record. The proposed framework contributes to the building of faster, data-driven disaster response systems in India and can be further enhanced with advanced language models, IoT and geospatial integration, and large-scale deployments with disaster management agencies.