Scaling Open Source ML: How Wolt Uses K8s To Deliver Great Food to Millions

Creating a machine learning platform that optimizes for both data metrics and business metrics is a challenge. The platform should be easy to launch and iterate, have common tooling and infrastructure, and prioritize automation and monitoring. The goal is to make machine learning a core business component.

• Disconnect between data metrics and business metrics is a challenge for data scientists
• Machine learning platform should optimize for both data metrics and business metrics
• Platform should be easy to launch and iterate
• Common tooling and infrastructure is important
• Automation and monitoring should be prioritized
• Machine learning should be a core business component

The speaker discussed the struggle of arriving at a meeting as a data scientist with lower mean squared error but not knowing what it meant for the business. The platform should prioritize both data metrics and business metrics to optimize for both. The speaker also mentioned a tool that converts YAML to Terraform, making it easier for data scientists to access S3 buckets without having to write all the Terraform code themselves.

Forecasting supply and demand, serving restaurant recommendations and predicting delivery times. These are just a few examples of how Machine Learning is being applied at Wolt. Now with over 12 million users, scaling the ML infrastructure has been a significant challenge. This talk will highlight those challenges and how they were addressed by building an end to end MLOps platform on Kubernetes. You'll learn about the open source frameworks that Wolt integrated, specifically Flyte, MLFlow and Seldon Core.Click here to view captioning/translation in the MeetingPlay platform!