Case Study: Scribd achieves ML platform standardization with Weights & Biases

Scribd standardizes ML workflows with Weights & Biases across 2 model lifecycle scenarios

Scribd, a provider of an eBook and audiobook subscription service, faced challenges with standardized machine learning practices across its disparate ML teams. Their ML Platform team, led by Staff Engineer Christian Williams, identified a need to reduce confusion and technical debt as their model landscape grew. They turned to the vendor Weights & Biases, specifically utilizing its experiment tracking and model management capabilities, to bring consistency to their workflows.

The solution involved integrating Weights & Biases Registry into a sophisticated CI/CD pipeline for model training and deployment. This provided critical versioning, aliasing, and lineage tracking, which helped break down team silos and streamline the process from experiment to production deployment on Amazon SageMaker. As a result, Weights & Biases provided a unified platform that reduced cognitive load for engineers and gave data scientists a rich toolset, leading to a high level of standardization and more effective model management across the organization.

Scribd