Needle bearings are fundamental components and widely used in automotive and industrial systems, ensuring smooth operation, efficiency, and longevity. Early detection of bearing defects during End-of-Line (EOL) testing and operational phases is crucial for preventive maintenance, thereby preventing system malfunctions. In the era of Industry 4.0, vibrational, accelerometer, and other IoT sensors are actively engaged in capturing performance data and identifying defects. These sensors generate vast amounts of data, enabling the development of advanced data-driven applications and leveraging deep learning models. Convolutional neural networks (CNNs) were mostly used in the earlier research to detect and identify bearing faults and more recent study showed the state-of-the-art Vision Transformer (ViT) achieving superior performance on identifying ball bearing defects. This paper introduces a novel approach to needle bearing defect detection leveraging the power of Contrastive Language-Image Pre-training (CLIP), an AI model capable of performing higher zero-shot learning capabilities. We detect needle bearing faults using CLIP fine-tuned model with spectrograms generated from Multitaper spectral estimation. Unlike traditional techniques, this approach leverages advanced Vision Transformer (ViT) embeddings that are pretrained with contrastive learning is fine-tuned with spectrograms for feature extraction, achieving robust classification performance. Results demonstrate high accuracy and robustness in fault detection, outperforming standard methods. Our research demonstrates the transformative impact of this AI-powered visual search system within a real-world manufacturing setting. This innovation has the potential to revolutionize needle bearing diagnostics, enabling proactive maintenance and preventing costly downtime in automotive systems.