Data science has already had an important effect on science, economy, and society, and its impact continues to grow as high-throughput technologies are being developed, business models adapt, and new opportunities for exploiting data for societal proposes are being developed. Today, the uptake of data science in practice is hindered mostly by a growing skills gap. Indeed, the application of data science techniques still requires sophisticated and rapidly evolving technical skills, and the demand for such skills grows faster than the suitably trained workforce. This project aims to address this skills gap by developing novel principles for automating data science tasks that today seem too openended, too domain-dependent, or involve data that is too complex, to allow doing so.