Digital, data-driven financial advisory
summary
Prototype asks for income and costs, hits Python models that take 7s (optimized to 1.5 seconds but still unacceptable), and collapses under load for 10 million customers. Solution is a real time pipeline: transaction labeling, Kappa architecture fed by Kafka streams, precompute insights on data change rather than on login. Major obstacles: heterogeneous sources, GDPR/compliance, need for federated data scientists, continuous data-quality monitoring. A constrained LLM pilot runs in production in the Czech bank with strong guarding and monitoring.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:07Serbia and Vienna data surprises
- 3:14Erste Group digital banking
- 6:29Digitalizing financial advisory
- 8:18First advisory prototype
- 11:55Data processing challenges
- 12:56Bank data complaints
- 17:17Data as a Service platform
- 18:47Real-time customer data
- 21:17Traditional BI failures
- 24:50Digital versus banking data
- 26:17Customer data patterns
- 27:22Kappa architecture
- 29:11Data insights in action
- 32:20AI challenges in banking
- 33:50Future of digital advisory
related talks
49mBuilding a Data-Driven Bank
46mMaking Data Relevant to Business. It's Harder Than You think!
13mTransforming Financial Services with Real-Time Analytics
01h 03mAI Innovation for a Smarter Financial Ecosystem
16mOperationalizing AI in Finance: Real-Time Analytics
31mBuilding Data-Driven Client Relationship Management in Banking with Python
29mBuilding Resilient AI Platforms: Financial Fraud Detection
33mSmart Banking - Real Time Driven