Learn Vertex AI while building a fraud detection system
summary
Builds a working fraud detection app called Fraud Finder that uses BigQuery for feature engineering, exports aggregated features into the Vertex AI Feature Store for point-in-time lookups, and deploys models as Vertex AI endpoints wired into pipelines and monitoring. The demo streams synthetic transactions, reads features at prediction time, and shows prediction latency in milliseconds and a dashboard of fraud-rate vs probability threshold. Time-aligned features prevent leakage, pipelines gate deployment by metric thresholds, and the system produces low‑latency online scores in the demo.
AI-written summary · about
chapters (4)
related talks
12mBuild end-to-end solutions with Vertex AI
01h 52mDeploying a Fraud Detection Model with a Feature Store
31mLet's do graph machine learning on Cloud
45mVertex AI: Pipelines for your MLOps workflows
30mAI in the Cloud
40mVertex AI: Pipelines for your MLOps workflows
31mGoogle's AI Map
06m[Flash talk] Fraud detection at scale using AI @ Razorpay