© 2018 The Author(s). During the last decade, a considerable amount of effort has been made to classify variable stars using different machine-learning techniques. Typically, light curves are represented as vectors of statistical descriptors or features that are used to train various algorithms. These features demand high computational power and can last from hours to days, making it impossible to create scalable and efficient ways of classifying variable stars automatically. Also, light curves from different surveys cannot be integrated and analysed together when using features, because of observational differences. For example, having variations in cadence and filters, feature distributions become biased and require expensive data-calibration models. The vast amount of data that will be generated soon makes it necessary to develop scalable machine-learning architectures without expensive integration techniques. Convolutional neural networks have shown impressive results in raw image classification and representation within the machine-learning literature. In this work, we present a novel deep-learning model for light-curve classification, based mainly on convolutional units. Our architecture receives as input the differences between the time and magnitude of light curves. It captures the essential classification patterns, regardless of cadence and filter. In addition, we introduce a novel data augmentation schema for unevenly sampled time series. We tested our method using three different surveys: the Optical Gravitational Lensing Experiment (OGLE-III), the Visible and Infrared Survey Telescope (VISTA) and Convection, Rotation and planetary Transit (CoRot) which differ in filters, cadence and area of the sky. We show that, besides the benefit of scalability, our model obtains state-of-the-art level accuracy in light-curve classification benchmarks.