The integration of data from smartphones and wearable devices offers a groundbreaking opportunity to apply machine learning for advancements in digital health. This project presents a case study demonstrating the application of advanced machine learning techniques to large-scale, heterogeneous datasets, with a focus on identifying clinically relevant biomarkers and enabling personalized therapeutic pathways. The project highlights the challenges inherent in managing and analyzing the complexity of physiological and environmental data streams, enriched by user annotations such as mood tracking and medication intake. By leveraging high-performance computing (HPC) infrastructures, the methodology addresses the heterogeneity, volume, and real-time requirements of these datasets. This poster will provide a detailed examination of how HPC-enabled workflows facilitate the preprocessing, feature extraction, and analysis of multi-modal data. It will also illustrate the scalability of the approach, offering insights into the translation of digital health data into innovative therapeutic interventions. The discussion will emphasize the computational techniques used, the challenges of HPC adaptation for machine learning, and the clinical relevance of the findings, while focusing on the scalability and reproducibility of the methodology.