Explore a 32-minute conference talk from MLOps World: Machine Learning in Production on quick deployment of model serving in ranking systems. Discover how Yelp improved their process for updating and replacing models in ElasticSearch by developing an ES plugin integrated with their Model Platform. Learn about the implementation of Spark Pipelines trained and stored as MLeap bundles to MLFlow, allowing direct deployment to ES as MLeap pipelines. Understand how this approach encapsulates both the ML model and its feature engineering pipeline, enabling seamless swapping of modeling pipelines. Gain insights into the ES plugin's development and valuable lessons learned in optimizing model pipeline performance for efficient ranking systems.
Quick Deploy Model Serving in Ranking Systems
MLOps World: Machine Learning in Production via YouTube
Overview
Syllabus
Quick Deploy Model Serving in Ranking Systems
Taught by
MLOps World: Machine Learning in Production
Related Courses
-
Building a Reproducible Model Workflow
-
MLOps Course – Build Machine Learning Production Grade Projects
-
End-to-End Machine Learning Project: From EDA to MLOps Integration
-
End-to-end machine learning operations (MLOps) with Azure Machine Learning
-
Perform data science with Azure Databricks
-
MLflow Model Serving - From Training to Production
