GS Caltex reports cutting model deployment from about a week to hours

GS Caltex says its machine learning engineers now deploy models within a few hours, compared with about a week previously, after automating the handoff from Amazon SageMaker Canvas to SageMaker Pipelines. An AWS technical post describes how the South Korean energy and chemical company reuses models and data preparation flows validated by process engineers. The company runs batch predictions every 30 minutes, reducing its need for continuously running inference endpoints, and reports meaningful inference cost savings. Engineers still configure production data paths, review model approvals and connect processing jobs. Scheduled runs also require checks to prevent overlapping executions and mismatched prediction results.