With rising mental health concerns, there is an urgent need for intelligent digital tools that provide empathetic and timely support. This paper introduces EmoTrax, a multi-modal emotion recognition system that leverages AI and NLP to detect emotional states in real time. It processes both text and voice inputs to generate personalized mental health recommendations using deep learning models. The system incorporates speech-to-text conversion and sentiment classification to ensure accuracy and contextual relevance. EmoTrax is designed with a user-friendly interface and a scalable backend for seamless interaction. Ethical design is a core focus, with safeguards for privacy, data sensitivity, and user autonomy. Continuous learning is supported through feedback loops and user interaction. The paper also highlights system limitations, such as potential bias in emotion detection, usability challenges, and reliance on third-party APIs. Evaluations of performance and usability show promising results, demonstrating high emotion detection accuracy and positive user engagement. Future developments include integrating facial recognition and physiological signals to enhance emotional insight. These enhancements aim to promote emotional awareness and early intervention. Overall, EmoTrax offers a scalable, responsive solution that bridges the gap between AI technology and mental health support. Keywords: AI, deep learning, digital wellness, emotion recognition, ethical AI, mental health, multimodal input, NLP, sentiment analysis, user interface.