Autism is one of the most prevalent and heterogeneous neurodevelopmental disorders; yet, despite decades of research, the neurobiology of autism is still poorly understood. The current paradigm for investigating the neurobiology of autism has reached a crossroads: inconsistent findings from underpowered studies fail to address the heterogeneity in autism, precluding the identification of robust and interpretable neurobiological markers and predictors of clinical symptoms, which can be used as more accurate diagnostic measures and precise treatment targets. In this talk, I will present our work toward addressing this challenge by leveraging newly available large-scale brain imaging and clinical data as well as exciting recent advances in explainable artificial intelligence methods.