Health Friend, an edge-based embedded system for ongoing cardiac monitoring and early identification of potentially lethal arrhythmias, is designed and developed in this study. To precisely record and digitize ECG signals, the system combines a Raspberry Pi 5, an AD8232 ECG sensor, and an MCP3008 ADC. A Butterworth low-pass filter is used to eliminate high-frequency noise and guarantee signal clarity. Using the MIT-BIH Arrhythmia Database as training data, a one-dimensional convolutional neural network (1D-CNN) divides heartbeats into five groups according to AAMI criteria. Real-time, cloud-independent ECG analysis is made possible by the model's local operation on the Raspberry Pi. Its usefulness in wearable, home care, and remote health monitoring applications is increased when detected anomalies cause alarms to be triggered by GPIO-driven actuators such buzzers or LEDs. This work demonstrates a scalable, cost-effective, and patient-centered approach to integrating embedded deep learning with biomedical signal processing for reliable cardiac health surveillance.