In this work, we present Berkeley Open MOS dataBase (BOMB), the first ever open source database of un-annotated transistor characterization data that can be used for technology representation learning. BOMB was created by running large scale transistor characterization scripts on various device flavors and technology nodes. Each data point in BOMB is essentially a multi-dimensional array capturing I-V and Y-V characterization of transistors. The meta-information revealing the technology is removed to maintain confidentiality such that the dataset to be used by a greater community of researchers. Additionally, we present an API with which the data can be accessed and visualized, as well as a framework with which additional data can be collected and added to the dataset. Finally, we provide statistics about the distribution of datapoints and visualizations which demonstrate the inherent structure of the dataset. Plans for future work involving downstream machine learning tasks are also discussed.