DeepStreaks identifies streaking fast-moving near-Earth objects in Zwicky Transient Facility (ZTF) image data using a convolutional neural network. The system applies a deep-learning classifier to candidate streak detections in ZTF images to separate genuine fast-moving objects from artifacts and other spurious streaks. It operates within the ZTF data-processing workflow to support high completeness in fast-moving object discovery while maintaining a low false-positive rate.