Description
Fraud now moves at machine speed, but most detection still runs on overnight batch jobs. In this demo, Razi Bayati, a Delivery Solutions Architect at Databricks, shows how a fraud analyst AI coworker detects telecom fraud in real time on one platform, from raw network data to a case-management workbench for analysts, managers and executives.
What you'll learn
• How every transaction is scored against network signals: impossible travel, cell and IP country mismatch, rapid cell-tower hops, roaming anomalies and device trust
• How clear fraud, like an international SIM swap, is auto-blocked, uncertain cases go to an AI agent and an analyst for review, and legitimate customers are approved with zero friction
• How Lakeflow, Lakebase, Unity Catalog, Genie and Databricks Apps fit together in a unified fraud architecture
• The Executive, Management and Analyst views of a custom Databricks Apps workbench, including Genie answers, a 30-day AI forecast of case volume, an AI investigation agent and a Knowledge Assistant trained on the fraud playbook
Chapters
0:00 Introduction and demo overview
1:17 The AI coworker in action
3:17 One platform, layer by layer
4:36 The executive view
5:49 The management view
6:32 The analyst view
Resources: The Fraud Analyst AI Coworker: https://www.databricks.com/blog/fraud-analyst-ai-coworker-detecting-telecom-fraud-real-time-one-platform