<p>Tool condition monitoring (TCM), powered by sensor technology and artificial intelligence (AI), has been adopted in the machining industry but faces issues such as data quality and model generalization. A classical transfer learning approach, where a pre-trained model trained on a large labeled dataset is fine-tuned to the target task, can mitigate model generality challenges. However, collecting abnormal data that represents faulty machining states is prohibitively expensive, making it difficult to gather sufficient, high-quality training data. Moreover, the limited computational resources on CNC machines complicate AI deployment. To address those problems, we develop DeepMachining, a deep learning-based AI system for real-time error prediction in lathe machine operations. We built and evaluated DeepMachining using real manufacturing data in practice. Specifically, we first pretrain a deep learning model to learn the representation of machining states. Then, we fine-tune it for specific machining tasks. The validation results show that DeepMachining provides high prediction accuracy for diverse workpieces and cutting tools. To the best of our knowledge, this work is one of the first industrial demonstrations of pre-trained deep learning models for predicting lathe machining errors.</p>