Microscopy is the oldest and most versatile technique for the study of cellular phenotype, but it has been underutilized for modern functional genomics and therapeutic discovery efforts due to traditional limitations on throughput. Recent advances have overcome these historical bottlenecks, however, and we have entered a new renaissance in microscopy, where innovations in imaging and machine learning have enabled the high-dimensional single-cell profiling of cellular images with unprecedented resolution and scale. Recently, we have developed combined in situ sequencing/multicolor labeling protocols enabling the creation of optical phenotyping pipelines of extraordinarily high throughput, at drastically reduced cost compared to existing single-cell profiling methods. By combining these pipelines with new machine learning approaches, we hope to unlock new insights about cancer vulnerabilities, enable therapeutic discovery, and elucidate the role of human genetic variation in health and disease.