eBay upgraded its AI infrastructure by adopting Ray, cutting model development and deployment time by roughly 50% and boosting GPU utilization from about 10% to over 75% through automated cluster scaling and high‑throughput batch inference.
This article details eBay's practical experience integrating multimodal data and graph neural networks into its recommendation pipeline, covering pain‑point analysis, a twin‑tower multimodal embedding model with triplet loss and TransH, engineering design, experimental results, and key takeaways for future AI‑driven product development.
This talk by eBay software engineer Zhu Feng explains the fundamentals of Delta Lake and Lakehouse architecture, outlines eBay’s migration from Teradata to a Spark‑based platform, and details the custom enhancements, performance optimizations, and operational improvements implemented to support large‑scale update and delete workloads.
This article details eBay's evolution of its massive HDFS storage—from a single‑cluster design to ViewFS Federation, then to Router‑Based Federation—highlighting the performance bottlenecks, optimization techniques, FastCopy integration, and future plans for further scaling and automation.
A Swiss EPFL master's student replaced the iPhone X's Lightning port with a USB‑C connector, sold the custom‑made device for over $86,000 on eBay, and sparked discussion about universal charging standards, Apple’s Lightning persistence, EU legislation, and the growing demand for a single charging cable worldwide.
This article explains how eBay uses graph neural networks and a heterogeneous‑graph fraud detection framework (xFraud) to improve payment risk assessment, overcome the limitations of traditional machine‑learning models, and effectively identify both individual and organized fraud in a large‑scale e‑commerce environment.
The article explains how eBay protects its global payment system using a graph‑neural‑network driven risk management framework called xFraud, which combines heterogeneous graph sampling, node‑type encoding, attention mechanisms and dynamic‑graph extensions to detect and explain both individual and organized fraud patterns in real‑time.
The article outlines eBay’s practical scalability principles—including functional partitioning, horizontal sharding, avoiding distributed transactions, aggressive asynchronous decoupling, moving work to async pipelines, virtualization, and intelligent caching—demonstrating how large‑scale web systems can be designed for linear growth and high availability.
The article explains eBay's use of MongoDB to create highly available, fault‑tolerant multi‑data‑center architectures, detailing design patterns, replica set configurations, read/write strategies, and recent MongoDB features that enable scalable, mission‑critical applications.
This article outlines eBay's seven scalability best practices—including functional partitioning, horizontal sharding, avoiding distributed transactions, asynchronous decoupling, streaming, virtualization, and smart caching—to help large‑scale web services achieve reliable, cost‑effective growth.
This article shares eBay’s seven core scalability best practices—including functional partitioning, horizontal sharding, avoiding distributed transactions, asynchronous decoupling, stream processing, virtualization, and smart caching—to help architects design highly available, cost‑effective systems that can handle billions of daily requests.
The article provides a comprehensive technical overview of Apache Eagle, an open‑source, distributed, real‑time security monitoring and alerting platform for Hadoop developed by eBay, covering its motivation, architecture, core components, machine‑learning based detection, typical use cases, and future development directions.
