- Abstract: Reproducibility of ML models and pipelines deployed in production requires capturing both the current and historic state. Capturing state in ML pipelines is complex due to the inherent nature of typical production deployments. We present a system that addresses these issues from a systems perspective, enabling ML experts to track and reproduce ML models and pipelines in production. This enables quick diagnosis of issues that occur in production.
- TL;DR: A systems perspective on reproducibility of machine learning models/pipelines deployed in production, that captures/tracks and versions the execution history along with its parameters and causality to enable quick diagnosis of issues.
- Keywords: reproducibility, systems, orchestration, machine learning, deployment