The Big Data Systems (BDS) group at UH focuses on developing interoperable, scalable and parallel algorithms to analyze data sets with machine learning (e.g. clustering,classification, regression, dimensionality reduction, variable/feature selection, time series and even deep neural networks), and graphs (connectivity, paths, cliques,vertex neighborhood). Our approach is unique in the sense that we build libraries, tools and systems in general, instead of using existing libraries. In other words, we aim to understand how analytic algorithms work all the way from reading the data set from secondary storage, processing it in main memory exploiting modern CPUs with a small RAM footprint, without reaching or exceeding the computer capacity. On the other hand, we develop interoperable routines that can exchange and read data in diverse formats, to analyze them with flexibility and high speed. In the past most of our research focused on parallel DBMSs and ''Big Data'' Hadoop systems. Today we shifted to modern data science languages like Python and R, where the data sources are diverse large files, which may not necesarily come from a relational database or data lake.