Risk Management
WeTech relies on large-scale user data to evaluate lending risk, but as its business expanded, big-data model training became a bottleneck. The company needed a mature and reliable data-science platform that could break through single-machine training limits, improve training efficiency, and avoid the staffing and maintenance burden of building a full-featured model-training environment from scratch.
Reported impact
WeTech replaced a heavier self-managed training approach with a managed lakehouse training stack that could scale with the business. The new platform delivered more than 3x engine performance versus the open-source baseline, second-level resource provisioning, and a complete migration of all big-data compute tasks to the cloud in roughly one month, while reducing overall storage and operations costs.
WeTech standardized its machine-learning platform on Alibaba Cloud EMR Serverless Spark and OSS. Instead of managing long-lived compute clusters, the team uses the serverless Spark service as a unified execution layer for data processing, analysis, and machine-learning jobs, taking advantage of the Fusion engine, vectorized execution, and RSS enhancements to improve performance over open-source Spark. The platform is tightly integrated with the broader lakehouse environment: dw-shell handles data collection, Apache Paimon provides the lake storage format, and Spark, Flink, and Hive are combined to form a full data-lake ecosystem for training and monitoring scenarios. By using compute-storage separation with OSS and task-level elastic provisioning, WeTech can run large-scale risk-model training with much lower infrastructure overhead while still giving data scientists familiar Python, SQL, PySpark, notebook, and scheduling workflows.
Collected lending and user-behavior data is gathered through WeTech's ingestion tooling into the lakehouse, stored through OSS and Paimon, and then processed by EMR Serverless Spark for feature preparation, model training, and analytical jobs. The resulting models and datasets are used by WeTech's financial risk-evaluation workflows and related real-time monitoring or analysis scenarios.
For lake storage, WeTech adopted Apache Paimon and combined it with Spark, Flink, and Hive to form a complete data-lake ecosystem that already supports real-time monitoring and analysis scenarios. (original: 在数据入湖方面，微财采用了 Apache Paimon 作为数据湖存储框架，并集成了 Apache Spark、Flink 和 Hive 作为计算引擎，构建了一个完整的数据湖生态系统。)
In the machine-learning scenario, the two core problems were breaking through single-machine training limits and improving training efficiency. (original: 在机器学习场景下需要解决的问题如下：一是要突破单机训练使用的数据规模的瓶颈。二是要提高训练的效率。)
WeTech chose EMR Serverless Spark because the managed platform unified data engineering and data science, and its Fusion engine, vectorized execution, and RSS capabilities stood out in evaluation. (original: 微财最终选择了阿里云 EMR Serverless Spark。作为一站式全托管湖仓分析平台，其自研 Fusion 引擎，内置高性能向量化计算和 RSS 能力，统一的数据工程和数据科学等，都是微财决定选择 Serverless Spark 的重要原因。)
WeTech depends on big data to evaluate risk during the lending process, and the growing scale of its datasets made model training a bottleneck. (original: 微财依托大数据评估用户借款过程中的风险，随着微财业务的快速发展，积累了大量用户数据，大数据集的训练逐渐成为瓶颈。)
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 page content states the company is 属小型企业（2024年参保人数33人）; the standard mapping classifies fewer than 250 employees as SMB.
Related
Alibaba Cloud Sample Archived · Jun 9, 2026 GTCFX As a fast-growing global brokerage, GTCFX required a robust and scalable IT infrastructure to support its expanding operations and increasing trading volumes. Key challenges included: Geographic Proximity: Ensuring data… Cloud Backup · Cloud Enterprise Network · Elastic Compute Service (ECS) Alibaba Cloud Sample Feb 21, 2025 Rongshan Technology (融山科技) Rongshan Technology was tackling highly fragmented personal-loan nonperforming assets with labor-intensive manual teams, low customer reach, and growing compliance exposure. The company needed a full-process AI approach… daily customer reach improved 20x Asset-management system · Intelligent dialog analysis · Large-model AI outbound system Alibaba Cloud Sample Dec 8, 2024 Shuhe Technology (数禾科技) Shuhe Technology had to run growing numbers of online risk models for pricing, limits, and loan-term decisions, but the infrastructure under those services could not scale fast enough with request volume. As inference… model-serving cost reduced about 60% ACK Serverless Knative · Elastic Container Instance (ECI) 