Energy and water related problems are becoming more relevant due to their huge impact on our environment. The limited availability of resources necessitates the development of machine learning prediction models that can help in predicting demand and consumption of these resources. We follow a data-driven approach that takes advantage of the data collected about the demand and usage of these resources. Our prediction models help in the decision making processes involved in the management of these resources. Our research focuses on developing deep learning and probabilistic models for sequential data generation and prediction. More specifically, we are focusing in water quality and availability prediction and in the energy disaggregation problem. In the following paragraphs we describe the problems, methods, data sets and some of the results of these ongoing projects. The models applied to these two problems can be extended to other smart living problem such as water demand and distribution, traffic prediction, and transportation demand.