Wayfair's Supply Chain Science team uses machine learning to provide predictive capabilities for use cases such as delivery-time prediction, incidence-costs prediction, or supply-chain simulation.
The team previously used Apache Airflow to orchestrate their models, experiments, and data pipelines. However, this caused several issues, including slow delivery processes and noisy neighbor problems.
To address these issues, the team migrated to Vertex AI Pipelines. Vertex AI Pipelines is a serverless, managed ML pipeline service that provides a number of benefits, including:
- Improved reliability and scalability
- Reduced time to value
- Increased reproducibility and traceability
The team also took the opportunity to refine their MLOps tooling and processes. For example, they now manage all of their pipelines centrally within a single model repository.
As a result of these changes, the team has seen significant improvements in their ML workflows. They are now able to develop and deploy models more quickly and easily, and they have greater confidence in the quality and reliability of their models.
The team is continuing to work with Vertex AI to further improve their ML capabilities. They are currently working on onboarding Vertex AI Feature Store and evaluating other Vertex AI services.
Here are some of the key point you can learn from the Wayfair's team experience:
- Vertex AI Pipelines can help to improve the reliability, scalability, and reproducibility of ML workflows.
- Centralizing ML pipelines can help to reduce complexity and improve collaboration.
- Adopting MLOps best practices can help to ensure the quality and reliability of ML models.
Click on the links below to read it more details.
Part-1
Part-2.
