The strategic necessity for autonomous safety systems has become paramount as traditional manual-trigger applications often fail during critical moments of panic or unconsciousness. This research presents the development of an AI-powered IoT wearable designed to provide proactive, hands-free protection. The objective was to engineer a TinyML-integrated device capable of identifying distress without user intervention, specifically addressing the gap where victims cannot physically reach a smartphone. The methodology centers on a sophisticated sensor fusion approach, utilizing Photoplethysmography (PPG) for heart rate monitoring, Inertial Measurement Units (IMU) for fall detection, and high-fidelity I2S microphones for voice-activated "wake word" and distress tone recognition. By shifting intelligence to the "edge" using the ESP32 microcontroller, the system performs real-time inference using an 8-bit quantized neural network, preserving privacy by processing audio locally without cloud transmission. Experimental results validate the system’s efficacy, demonstrating a 92% accuracy rate in AI-driven emotion classification and an emergency alert delivery speed of less than 5 seconds. The integration of GPS and GSM modules ensures precise location tracking within a ±5-meter range, functioning independently of cellular data. This research signifies a critical shift from reactive to proactive safety technology, providing a reliable, discreet, and automated guardian that significantly enhances user confidence and contributes to a broader societal framework for public safety.