RLOps Pipeline Development with Low-Code and Large Language Models for Industry 4.0: Replication Package

Title: RLOps Pipeline Development with Low-Code and Large Language Models for Industry 4.0: Replication Package Authors: Stephen John Warnett; Uwe Zdun About: This is the replication package artefact for the paper entitled "RLOps Pipeline Development with Low-Code and Large Language Models for Industry 4.0". Paper Abstract: Machine learning operations (MLOps) automate common tasks throughout the entire life cycle of a machine learning model. In Industry 4.0 environments, cyber-physical production systems increasingly utilise reinforcement learning (RL) models to optimise and automate production processes, giving rise to reinforcement learning operations (RLOps). However, manually configuring RLOps pipelines requires specialised development operations (DevOps) knowledge, and this can limit the potential for automation. We propose a template-based approach for rapidly creating and deploying RLOps pipelines in Industry 4.0 environments that minimises the required programming effort and expertise. Based on the Pipes and Filters pattern, our modular solution leverages large language models (LLMs) for automated pipeline creation. It enables fully automated execution, including model training, testing and deployment, with built-in quality control to ensure correct configurations. We validate our approach through evaluation with multiple LLMs in an Industry 4.0 context. Our results demonstrate that our solution, used with a suitable LLM, can reliably generate and execute RLOps pipelines with low error rates, thereby reducing development time and the need for specialised DevOps knowledge.