In this session, we discuss some of the key challenges and opportunities currently faced by Data Science practitioners. We first consider the process of “deconstructing” data science questions, and illustrate with examples the extent to which slight modifications to questions may result in opposite conclusions from data. We follow with a discussion on machine learning and interpretability in the context of highly regulated verticals, such as banking and insurance. We then discuss optimization of data science pipelines with key focus on Bayesian Optimization for hyper-parameter search and other interesting applications. Finally, we introduce Counterfactual Inference methods, and discuss the extent to which they provide the right framework for how we reason about many problems faced by industry.