Scaling ML Embedding Models to Serve a Billion Queries

Explore the challenges and solutions in scaling embedding-based visual search at eBay in this 39-minute conference talk from MLOps World: Machine Learning in Production. Join Senthilkumar Gopal, Senior Engineering Manager, and Deepika Srinivasan, Senior MTS, from eBay's Search ML team as they delve into the intricacies of powering visual search at scale. Gain valuable insights into the model architecture, application architecture for serving users, and workflow pipelines developed for building embeddings used by Cassini, eBay's search engine. Discover the unique challenges encountered during the implementation process and learn key strategies for handling embeddings and scaling systems to provide real-time clustering-based solutions. This talk offers a comprehensive overview of the complexities involved in serving a billion queries through ML embedding models, making it essential for professionals working on large-scale machine learning applications in production environments.

Scaling ML Embedding Models to Serve a Billion Queries