KartShoppe：使用 Ververica 和 Apache Flink® 进行实时特征工程.pdf

1、 2025 VervericaKartShoppe:Real-Time Feature Engineering with Ververica and Apache FlinkKarin Landers-Sr.Product Marketing Manager 2025 VervericaHello,KartShoppe!2025 VervericaData=Business Growth 2025 VervericaKartShoppes Challenges 2025 Ververica Loyalty Is Not Guaranteed Its easier to swap apps th

2、an find a new brick and mortar store Keep up or get left behind Digital-Native Users Everything is online Massively increased attack vectorsChanging Customer Behavior Mobile-First User Experience Low bandwidth,low security devices Small screens,and shorter interactions 2025 VervericaAre We Keeping U

3、p?2025 VervericaGo FASTER!2025 VervericaGo FASTER SMARTER!2025 VervericaBuilding a Real-Time AI Pipeline 2025 VervericaData Ingestion Capture the event steam,regardless of data source 2025 VervericaData PreprocessingClean data=quality data 2025 VervericaFeature EngineeringTransform raw data into mea

4、ningful outcomes 2025 VervericaModel Serving and Inference Feeding the model 2025 VervericaAdaptive LearningAI models need to change and update without downtime 2025 VervericaFeedback LoopsOngoing model performance evaluations-learn by doing 2025 VervericaHow does Ververica help solve this?2025 Verv

5、ericaValue of ActionsMinutesMillisecondsSecondsTomorrows AI Fueled by Real-Time IntelligenceSmarter Decisions,Faster ActionsMore Accuracy,Better ExperiencesLower Costs,Always the Best DataPrevents Losses,Drives Growth 2025 VervericaUnified Streaming Data PlatformConnect to dozens,hundreds,or thousan

6、ds of Freshand Historic data sources 2025 VervericaConfidential-for internal use onlyFocus on Feature Engineering Actionable Signals Improved Model Accuracy Reduce Complexity More Relevant 2025 VervericaConfidential-for internal use onlyTime-Based Joins for Event Correlation 2025 VervericaConfidenti