This thesis explores two lines of research in the context of sequential data and machine learning in the remote environment, i.e., outside the lab setting - using data acquired from wearable devices. Firstly, we explore Generative Adversarial Networks (GANs) as a reliable tool for time series generation, imputation and forecasting. Secondly, we investigate the applicability of novel deep learning frameworks to sequential data processing and their advantages over traditional methods. More specifically, we use our models to unlock additional insights and biomarkers in human-centric datasets. Our first research avenue concerns the generation of sequential physiological data. Access to physiological data, particularly medical data, has become heavily regulated in recent years, which has presented bottlenecks in developing computational models to assist in diagnosing and treating patients. Therefore, we explore GAN models to generate medical time series data that adhere to privacy-preserving regulations. We present our novel methods of generating and imputing synthetic, multichannel sequential medical data while complying with privacy regulations. Addressing these concerns allows for sharing and disseminating medical data and, in turn, developing clinical research in the relevant fields. Secondly, we explore novel deep learning technologies applied to human-centric sequential data to unlock further insights while addressing the idea of environmentally sustainable AI. We develop novel deep learning processing methods to estimate human activity and heart rate through convolutional networks. We also introduce our ‘time series-to-time series GAN’, which maps photoplethysmograph data to blood pressure measurements. Importantly, we denoise artefact-laden biosignal data to a competitive standard using a custom objective function and novel application of GANs. These deep learning methods help to produce nuanced biomarkers and state-of-the-art insights from human physiological data. The work laid out in this thesis provides a foundation for state-of-the-art deep learning methods for sequential data processing while keeping a keen eye on sustain- able AI.