The growing incidence of cyber threats, identity theft, and unauthorized system access has highlighted the limitations of traditional authentication methods such as passwords and PINs. In response, this study presents the design and implementation of a Multi-Modal Biometric Authentication System on Amazon Web Services (AWS) using Deep Learning Techniques, integrating facial recognition and voice biometrics to enhance security and reliability. The system leverages cloud-native services for scalability and real-time processing, utilizing Amazon Rekognition for facial recognition and a deep learning-based speaker verification model, ECAPA-TDNN (Emphasized Channel Attention, Propagation and Aggregation Time Delay Neural Network) deployed on AWS SageMaker for robust voice authentication. Additional AWS services including Lambda, API Gateway, Amazon S3, and DynamoDB are used to orchestrate data processing, storage, and communication between system components. A comprehensive dataset comprising facial images and voice recordings from 100 participants was collected under diverse environmental conditions to ensure variability and robustness. The system processes biometric inputs uploaded via a React-based frontend, where facial and voice features are extracted and matched against stored templates. Experimental evaluation was conducted using a dataset of 1000 samples, with 800 samples used for training and 200 for testing. Performance was assessed using standard biometric evaluation metrics, including accuracy, False Acceptance Rate (FAR), and False Rejection Rate (FRR). Results demonstrate that while individual modalities achieve high performance, the fusion of face and voice in a dual authentication framework significantly improves overall system accuracy and reduces error rates. The study concludes that multi-modal biometric authentication offers a more secure and resilient alternative to unimodal systems by mitigating issues such as spoofing, environmental variability, and single-point failure. The integration of the ECAPA-TDNN model with AWS cloud services ensures efficient, scalable, and real-time authentication suitable for deployment in high-security domains such as banking, e-governance, and enterprise access control systems. This research contributes a practical and scalable framework for implementing advanced biometric authentication systems in modern cloud environments.