Machine learning allows uncovering and relating patterns of interest. This talk will highlight a set of research projects in our group that target mental health applications of machine learning using brain imaging and other phenotypic data. In recent years, we have witnessed an assortment of software and data that enables easy application of machine learning technology. However, several caveats need to be kept in mind when applying such technologies in the context of limited data and our typical iterative research processes. We will also present current and upcoming technology platforms, software, and data collections that enable researchers to minimize errors, and also to close and shorten the feedback cycle in research. While many of these technological advancements have transformed our group's research approaches and questions, their limited exposure across the domain of neuroscience indicates a large gap in the current education of neuroscientists in connecting their questions to relevant data and tools. To counter this deficit, we have been working on a set of collaborative initiatives to create resources that support the entire life cycle of experimental research and promote the importance of making scientific research "open by design".