Toward an integrated and supported machine learning process

In recent years, there has been a significant increase in the use of machine learning models by industrial companies with the aim of predicting potential failures in their machinery or obtaining data related to this. Creating, training, and storing machine learning models can be a complex process due to the considerable number of tools, technologies, and services available, which generates considerable debate in the industry. To address these issues, this paper proposes and explains how technologies such as Docker, MLflow, AWS and others support the industrial process by designing and deploying a runtime architecture for real use cases. These tools could help in the creation and deployment of ML models. The proposed solution includes scenarios such as remote servers deployed in the cloud or local servers managed directly by the company.