Federated AI in Future Digital Banking
summary
WeBank scales to 115 million users in four years with 2,000 employees and 50% technical staff, launching financial products in 11 days and running ~200 risk models for real‑time scoring. Vertical and horizontal federated learning (encrypted metadata, secret‑sharing planned) improves SME loan model accuracy by about 12% and cuts stress/failure cases ~40%. Open‑source FATE is donated to the Linux Foundation, ships FATE Flow, FATE Board and federated serving, and integrates with Kubernetes/Harbor, Kubeflow and Angel.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 0:41Where is WeBank
- 2:57Why digital banking is blooming
- 4:22Chinas massive construction projects
- 6:05unprecedented consumer market on the mobile internet
- 7:10digital by AI technology
- 8:23customer identity
- 9:02risk evaluation
- 10:00AI dream experience
- 10:37GDPR
- 12:40Data Regulations
- 13:50Predator Learning
- 18:53Exchangeable Encryption
- 19:24Machine Learning
- 22:09Use Cases
- 23:00Digital Invoice
- 23:44Model Accuracy
- 24:45Project
- 25:30Roadmap
- 27:59Project Landscape
- 28:10Core Component
- 29:19Submodels
- 29:44Model Version Control
- 30:00Online Feature Engineering
- 32:19Collaborations
- 33:03Harbor Project
- 33:23Angel Mission
- 34:13Resources
related talks
26mThe AI-Based Banking Era
01h 03mAI Innovation for a Smarter Financial Ecosystem
42mFederated Learning: Unlocking the Values of Discrete Data for Enterprise
24mFintech-Speed Banking: AI Blueprint
21mWhat, Why and How of Federated Machine Learning – Implementation Using...
25mData collaboration with Federated Learning
41mAI For All: Enhancing Digital Financial Services using Data Science
16mMy Bank Meets AI: Smarter Decisions Beyond Fancier Spreadsheets