Particle accelerators are arguably the most complex scientific instruments ever built. Given the enormous amount of data generated by these facilities, and the increasing demands on system performance, we recognize the need to leverage advances in machine learning (ML). In the last few years there have been several ML projects at Jefferson Lab with a common focus to optimize the operation of superconducting RF (SRF) cavities. This multi-faceted approach aims to (1) identify and classify types of faults from C100 cavities, (2) extend the work to provide real-time fault prediction for C100 cavities, (3) minimize radiation levels due to field emission in the linacs, and (4) develop tools to automate cavity instability detection in legacy cryomodules. We give a brief description of each project along with model performance. We also highlight challenges in working with real-world data and challenges for deploying models.