Autism Spectrum Disorder (ASD) is a developmental condition that affects communication, behavior, and social interaction. Meltdowns, common in people with autism, are intense emotional outbursts triggered by overwhelming sensory or emotional experiences, often requiring specific coping strategies for effective management. This study presents an advanced system for autism detection and personalized management, leveraging machine learning and computer vision for real-time monitoring and tailored rehabilitation of individuals with ASD. Our system integrates a Support Vector Machine (SVM) for initial autism detection, achieving an accuracy of 78.3%, with a novel Convolutional Neural Network (CNN) model that attains 93.3% accuracy in emotion recognition for meltdown. While the SVM highlights opportunities for refinement in feature selection, the CNN demonstrates effective classification of emotional states across facial expressions. Through correlation analysis, we identify key ASD predictors, including family history, social responsiveness, and developmental delays. Real-time meltdown detection is achieved by analyzing hand movements and emotional cues, generating adaptive rehabilitation suggestions that respond to the patient’s current needs.