As encrypted network traffic becomes increasingly prevalent, cyber network operators are operating in the dark with regard to the kinds of traffic flowing through their networks. Machine learning (ML) techniques have recently emerged that can rapidly learn and provide contextual labels to encrypted traffic. However, as ML-based applications reach the hands of operators, they do not always understand the limitations of the underlying models and their predictions, as ML models often struggle or fail to communicate the confidence of their predictions. Without this nuanced understanding, operators may blindly trust a model's predictions, unaware that the model is only marginally confident or has never seen the input data during training. QUETAL is a software prototype that puts ML into the hands of cyber analysts to (1) enable them to train and deploy models where existing tools cannot make any predictions, and (2) provide contextualized uncertainties with each prediction that! allow analysts to filter visualizations according to their desired confidence level and understand where the model may be inaccurate.