• -> DSBox: End-to-End Automated Machine Learning (E2E-AutoML) system
• -> Supports high dimensional structured datasets (image, text, video, etc.)
• -> Research on a Monte-Carlo Tree Search(MCTS) Based Scheduling algorithm

1. Qasemi E, Stan S, Yao K, Shao R, Liu J, Liang M, Ferrer LJ, Szekely P, "DSBox: Data Scientist in a Box", ISI Graduate Student Symposium (GSS '19), Marina Del Rey, California, USA.

End-to-End Automated Machine Learning (E2E-AutoML) seeks to automate the machine learning pipeline generation process by augmenting, cleaning, and featurizing arbitrary data, as well as parametrizing and tuning machine learning algorithms for optimal performance. Limitations in the media/task types these systems support, lack of support for an open set of machine learning (ML) primitives, and the black-box nature of the produced ML pipelines have made it difficult for such systems to be fully exploited by users across many expertise levels. In this work we present DSBox, an expandable, all-in-one E2E-AutoML system, that can pave the way to human-centric AI by meeting the needs of Non-ML experts, while also allowing seasoned Data Scientists to easily incorporate their knowledge through highly customizable templates that configure the pipeline search-space. We show DSBox is able to outperform human experts and beat state-of-the-art AutoML systems (ML-Plan, Auto-sklearn, TPOT) through evaluation on 468 diverse datasets.