The NetEase Yanxuan Technology Product Team shares practical tech insights for the e‑commerce ecosystem. This official channel periodically publishes technical articles, team events, recruitment information, and more.
This article details the motivation, design goals, architecture, and implementation steps of a reverse transaction microservice that handles cash and promotional refunds, reduces technical debt, and improves operational efficiency for a fast‑moving e‑commerce platform.
Yanxuan built a Kubeflow‑based machine‑learning platform that unifies data preprocessing, feature engineering, model training, validation, and deployment, using Smart‑jobs, Smart‑Infer, Smart‑backend, Airflow pipelines, Jupyter notebooks, and Istio‑enhanced inference services to boost algorithm engineers’ efficiency and integrate with Kubernetes, HDFS, and Hive.
This article details the design and implementation of a transparent, explainable traffic‑decision system for an e‑commerce platform, covering background challenges, directional principles, selection and targeting methods, PV value estimation, allocation algorithms, and the supporting data‑engineering and visualization infrastructure.
The article details a step‑by‑step evolution of an e‑commerce platform’s top‑slot recommendation system, moving from a DCN‑mix single‑objective model through BST‑based dynamic features, position‑bias debiasing, multi‑task MMoE learning, and finally BST with target‑attention, each yielding measurable CTR, conversion, and user‑value gains.
After three years of development, NetEase Yanxuan has open‑sourced OSSA, a TypeScript‑based, Taro‑powered UI component library that delivers over 30 cross‑platform components for WeChat, Alipay, ByteDance mini‑programs and H5, featuring monorepo management, on‑demand imports, automated testing, and a community‑driven roadmap.
This article examines the technical challenges of traditional e‑commerce live streaming, presents a 24‑hour AI‑driven virtual live‑broadcast system built by Yanxuan and NetEase Fuxi, and details its architecture, virtual‑human generation, automated content creation, intelligent interaction, and future expansion plans.
The article explains how space‑efficient probabilistic structures such as BloomFilter and Count‑min Sketch enable large‑scale data deduplication, join pruning, real‑time idempotent filtering, and approximate top‑K analytics by trading modest accuracy loss for dramatically reduced storage and faster computation.
This comprehensive guide walks you through the entire front‑end performance optimization lifecycle—from understanding why performance matters and setting measurable goals, to building metrics, implementing concrete strategies across caching, networking, rendering, and finally validating results with monitoring and retrospectives.
The article presents a comprehensive end‑to‑end traffic data architecture for an e‑commerce platform, detailing hybrid frontend/backend tracking with SPM/SCM/action standards, data‑warehouse construction of fact and dimension tables, UUID i_code unification, real‑time attribution methods, and future automation of warehouse and model layers.
The article presents an Indicator Testing Platform that automates metric validation—covering timeliness, completeness, accuracy, and consistency—through model‑level comparison, regression, online monitoring, and TDD‑style testing, dramatically reducing manual effort and enabling rapid detection and correction of data quality issues across thousands of business indicators.
