As a result of their ability to amplify input light, ultra-high quality factor (Q) whispering gallery mode optical resonators have found numerous applications spanning from basic science through applied technology. Because the Q is critical to the device’s utility, an ever-present challenge revolves around maintaining the Q factor over long timescales in ambient environments. The counter-approach is to increase the nonlinear coefficient of relevance to compensate for Q degradation. In the present work, we strive to accomplish both, in parallel. For example, one of the primary routes for Q degradation in silica cavities is the formation of water monolayers. By changing the surface functional groups, we can inhibit this process, thus stabilizing the Q above 100 million in ambient environments. In parallel, using a machine learning strategy, we have intelligently designed, synthesized, and verified the next generation of small molecules to enable ultra-low threshold and high efficiency Raman lasing. The molecules are verified using the silica microcavity as a testbed cavity. However, the fundamental design strategy is translatable to other whispering gallery mode cavities.