JointBERT implements the BERT-based model from the 2019 paper for Joint Intent Classification and Slot Filling. The model processes text through a BERT encoder to simultaneously predict the user intent and fill slot labels (e.g., identifying entities like locations or dates) in a single forward pass, optimizing a combined loss function. It supports optional CRF layers for enhanced sequence labeling and was evaluated on the ATIS and Snips benchmark datasets.