Run Real-time Predictions with an Inference Pipeline