Traditional smart manufacturing emphasizes data-driven decision making for individual systems. In a smart manufacturing network, computation services are widely used to transform data streams to decisions in heterogeneous manufacturing conditions and integrate human expertise for cyber-human collaborative tasks in various contexts. These computation services can adaptively improve productivity, quality, and flexibility of manufacturing to personalization. Such computation services in a smart manufacturing network poses significant challenges in selection of computation pipelines with limited communication and computation resources. Motivated by these challenges, this presentation will mainly discuss how to recommend contextualized computation services and pipelines in a smart manufacturing network with different manufacturing processes. A recommender system-based framework is proposed to support efficient computation pipeline selection, considering the reliability and responsiveness of the ubiquitous communication and computation resources. Manufacturing case studies were performed to validate the proposed methodology.