Marketing Personalization · Real-time Analytics
Zhiyi Technology needed to keep a fashion-intelligence platform responsive while running large-scale image-recognition, vector retrieval, trend analysis, and recommendation workflows. Its self-built CDH environment was complicated to operate, on-premises GPU servers created batch-recognition bottlenecks, and the business needed a data platform that could support visual search and fashion-trend insight at scale.
Reported impact
Alibaba Cloud gave Zhiyi a more operable and scalable fashion-intelligence base. The company reduced the operational burden of self-built Hadoop infrastructure, sped up large-scale recognition and retrieval workflows, and kept image-derived trend signals available for products that support style discovery, intelligent recommendation, and downstream supply-chain planning across thousands of fashion brands.
Zhiyi Technology evolved its apparel-intelligence stack on Alibaba Cloud by moving heavy compute from a self-built CDH cluster into MaxCompute and DataWorks while using managed Elasticsearch with the Proxima vector engine for interactive analysis and image-based retrieval. Product and social-fashion images are labeled and converted into 256-dimensional vectors by self-developed computer-vision models trained on GPUs that pull annotated images from OSS. Those vectors are then used in visual-search and trend-analysis workflows, while DataWorks handles orchestration and MaxCompute runs the large-scale batch computations that would have been expensive to keep in the old CDH environment. Elasticsearch supports ad hoc search and multidimensional analysis for trend products such as Zhikuan, helping designers and brands explore style trends, discover hot items, and connect image-derived signals to downstream flexible supply-chain decisions.
Fashion images and related product data are labeled and stored in OSS, used to train vision models that generate 256-dimensional vectors, and then fed into Elasticsearch plus Proxima for image retrieval and interactive analysis. DataWorks schedules the workflows and MaxCompute handles the heavier batch computations that power fashion-trend, recommendation, and design-support features.
Its core visual-search scenario converts each image into a 256-dimensional feature vector and uses Proxima for efficient vector indexing. (original: 将每一幅图抽象成高维（256 维）特征向量，然后将所有特征借助 Proxima 构建成高效的向量索引。)
The earlier self-built CDH cluster was complex to operate, and on-premises GPU servers became a bottleneck for large batch image-recognition jobs. (original: 自建 CDH 集群运维比较复杂... 一次批量识别图片的工作量很大，线下的 GPU 服务器计算性能有瓶颈。)
To reduce operating complexity, Zhiyi migrated compute tasks to MaxCompute and used DataWorks for orchestration and scheduling. (original: 为了降低运维复杂度，知衣科技将计算任务迁移到 MaxCompute，直接基于 DataWorks 进行任务编排调度。)
Zhiyi built a fashion-intelligence SaaS platform around image recognition, data mining, and intelligent recommendation for trend discovery and style suggestion. (original: 知衣凭借图像识别、数据挖掘、智能推荐等核心技术能力... 为服装企业和设计师提供流行趋势预测、设计赋能、款式智能推荐等核心功能。)
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
The official page lists '员工数量：约400人（2021年）'.
Related
Alibaba Cloud Sample Oct 13, 2026 TOPSPORTS (滔搏) TOPSPORTS wanted to scale enterprise AI agents beyond isolated pilots, but every new agent risked recreating the same knowledge, data, tools, and skills in separate silos. Business-analysis, membership, and marketing… business analysis time reduced to 20 minutes AgentCore · AgentLoop · MCP tools Alibaba Cloud Sample Nov 14, 2024 Lenztech (Trax China) (朗镜科技) Lenztech had to process millions of retail shelf images and massive SKU catalogs every day while maintaining very high recognition accuracy and low latency. Complex in-store scenes, frequent model updates, and the need… 95%+ of complex flows finished within 10 seconds CPFS · Cloud-native AI suite · DeepGPU 
Google Cloud Sample Modified · Oct 9, 2026 Central Retail Food Replaced ~2,000 monthly manual Excel reports with a single source of truth on BigQuery through Looker dashboards Central Retail Food unifies its ecosystem on Google Cloud, moving from fragmented data to conversational… Addresses 50% to 60% of customer service queries with automated chatbots built on Conversational Agents BigQuery · Gemini · Gemini platform 