As data-size is increasing exponentially, new paradigm shifts have to emerge allowing fast exploitation of data by every- body. Large-scale predictive analytics is restricted to wealthy organizations as small-scale enterprises (SMEs) struggle to compete and are inundated by the sheer monetary cost of either procuring data infrastructures or analyzing datasets over the Cloud. The aim of this work is to study mechanisms which can democratize analytics, in the sense of making them affordable, while at the same time ensuring high efficiency, scalability, and accuracy. The crux of this proposal lies in developing query-driven solutions that can be used off the Cloud thus minimizing costs. Our query-driven approach will learn and adapt on-the-fly machine learning models, based solely on query-answer interactions, which can be used for answering analytical queries. In this abstract we describe the methodology followed for the implementation and evaluation of the system designed.