Risk Management
Wuxi Rural Commercial Bank needed to turn its historical data into practical AI risk capability, but it lacked deep internal AI talent and could not afford a long, high-cost platform build for machine-learning use cases such as credit assessment.
Reported impact
The bank established a fast, low-overhead machine-learning foundation for credit risk decisions, improved recall without reducing approval quality, and created a deployable AI pattern it can extend into future fraud and credit-limit use cases.
Baidu AI Cloud and Wuxi Rural Commercial Bank deployed an AI training appliance that connected the bank's internal big-data platform with external third-party data and separated development, testing, and production workflows. In production, upstream external and internal data are cleaned and processed on the bank's big-data platform. In the development and test zone, those curated datasets are extracted into the AI appliance to train and repeatedly refine credit models. Once validated, the models are exposed back to the bank's loan and admission systems through automated prediction interfaces. This arrangement let the bank stand up a full machine-learning loop quickly: data preparation, model training, iterative tuning, deployment, and real-time scoring all sit on one controlled stack. The first live use case focused on credit access and rating, giving the bank a practical template it can later extend into anti-fraud and credit-limit scenarios.
Third-party external data and the bank's internal data are combined and processed on the big-data platform, then exported into the AI appliance for training and tuning. Once the credit models are validated, they are published back to the bank's business systems, where loan workflows call the scoring interface for automated predictions in production.
A model was successfully built within two weeks (original: 在短短2周的时间内，成功搭建出AI一体机模型).
After tuning 60,000 samples, recall improved 50% at the same pass rate as expert rules (original: 通过对6万条样本信息的反复测算和模型调优之后…在与专家规则同等通过率的前提下，召回率提升了50%).
External third-party data was combined with internal big-data-platform data before training models on the appliance (original: 将前置机的外部第三方数据与行内大数据平台中的数据相结合…将大数据平台的数据抽取到AI一体机训练模型).
The AI training appliance was the first deployed in the rural commercial banking sector and solved the bank's limited AI talent problem (original: AI训练一体机应用是农村商业银行业内首个落地的一体机，有效解决了农商行人才不足、AI能力有限等问题).
The bank plans to extend the project into anti-fraud and credit-limit scenarios (original: 未来…深入反欺诈、授信额度等领域的AI应用).
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.
Related
Baidu AI Cloud Sample May 14, 2025 Penghua Fund Management Co., Ltd. (鹏华基金管理有限公司) Penghua Fund needed to reduce manual work in research, operations, and compliance tasks such as classifying financial events, summarizing announcements and meetings, checking fund-style drift, and answering internal… Fund managers save more than 30% of the time previously spent gathering market intelligence. BaiGe Heterogeneous Compute Platform · Base Zhihui Agent (Financial APaaS Platform) · Internal Knowledge Base Baidu AI Cloud Sample May 14, 2025 Unnamed Joint-Stock Commercial Bank (某股份制商业银行) The bank needed one enterprise large-model foundation that could serve marketing, risk control, document Q&A, coding, and customer-service scenarios instead of leaving AI capabilities fragmented across separate teams… Annual engineering labor costs could fall by more than 20% with Wenxin Kuaima. Heterogeneous Model Service Cluster · Intelligent Customer Service Assistant · Knowledge Retrieval Middleware Baidu AI Cloud Sample May 6, 2025 Postal Savings Bank of China Postal Savings Bank of China wanted a bank-wide AI PaaS that could coordinate model development, training, and deployment across risk, marketing, operations, and document workflows instead of leaving AI capabilities… Approval efficiency improved 30x. BaiGe Heterogeneous Compute Platform · Baidu AI Middle Platform Solution · Postal Savings Brain 