Vendors participating in tenders face significant challenges in creating timely and accurate order quotations from Request for Order (RFQ) documents. The success of their bids is heavily dependent on the speed and precision of these quotations. A key bottleneck in this process is the time-consuming task of identifying relevant products from the product catalogue that align with the RFQ descriptions. We propose the implementation of an automatic classification system that utilizes a context-aware language model specifically designed for the electrical engineering domain. Our approach aims to streamline the identification of relevant products, thereby enhancing the efficiency and accuracy of the quotation process. However, an effective solution must be scalable and easily adjustable. Thus, we present a machine learning operations (MLOps) architecture that facilitates automated training and deployment. We pay particular attention to automated pipelines, which are essential for the operation of a maintainable ML solution. Furthermore, we outline best practices for creating production-ready pipelines and encapsulating data science efforts. Schneider Electric currently operates the solution presented in this paper.