Challenge and Motivation

• AI holds great potential for the optimization of production
• The use cases here are numerous (e.g. predictive quality)
• However, many developed models do not make it into production
• Reasons for this are both the complexity of deploying machine learning models and a lack of structured guidelines (e.g., in the form of components) on how to approach deployment

Objective

• Development of a component-based deployment guideline
• Realization of the guideline on the basis of concrete use cases from the community

Procedure

• Analysis of the deployment requirements and identification of suitable use cases from the community
• Derivation of the various guideline components such as deployment design, productionizing & testing, monitoring and retraining
• Validation of the guideline by application to identified use cases