– eBay’s marketplace has over 1.9 billion listings from millions of sellers, and uses ML, AI-driven search capabilities, and NLP to enhance user experience.
– By implementing Google Cloud Vertex AI and Vector Search, eBay accelerated engineering velocity and launched new features to reduce friction for both buyers and sellers.
– eBay collaborated with Google to bridge the gap between on-premises and public cloud, exploring Vertex AI and Vector Search to improve ML workflows and rapid model iteration for a more intuitive recommendations engine.
eBay’s marketplace features over 1.9 billion listings from millions of sellers, offering a wide range of products. Through the use of machine learning, AI-driven search capabilities, and natural language processing, eBay has improved the user experience by making it easier to find relevant products on their platform. By implementing Google Cloud Vertex AI and Vector Search, eBay has accelerated engineering velocity, allowing for faster iteration and the launch of new features to reduce friction for both buyers and sellers.
As a leading online marketplace, eBay has developed a robust in-house infrastructure for machine learning processes using advanced solutions such as an ML GPU cluster and Vector Similarity Engine. By partnering with Google Cloud to augment their ML workflow, eBay has been able to leverage advanced AI capabilities, scalability, and a reliable infrastructure. Vector Search has been particularly beneficial for generating deep-learning-based semantic embeddings for eBay listings, allowing for rapid iteration and quicker access to products for customers.
To enhance their recommendations engine, eBay’s Recs team built a custom model on Vertex AI and indexed their embeddings into Vector Search, enabling them to iterate and test the system’s effectiveness. With a hybrid interconnect link established between eBay’s data center and Google Cloud, eBay was able to run an A/B test comparing the performance of their in-house system with their Google Cloud infrastructure. While the real-time performance was similar, the onboarding process was much smoother and faster with Google Cloud. Through collaboration with Google’s network infrastructure team, eBay was able to create a framework that met their team’s and customers’ needs, ultimately improving the recommendations experience for users.