Building a Modern FinTech Big Data Infrastructure

发布于

3676

人观看

The cloud is now the first choice for large-scale analytics, but organizations that have sunk investment into Hadoop on-premises are also challenged with maintaining operations. This can make a move to modern analytics platforms like Spark difficult or impossible. Learn about innovations for large-scale migration that can take full advantage of cloud-based analytics without disrupting operations.

展开查看详情

1 .WIFI SSID:Spark+AISummit | Password: UnifiedDataAnalytics

2 .Building a Modern FinTech Big Data Infrastructure Omar Hommos & Rodel van Rooijen, Adyen #UnifiedDataAnalytics #SparkAISummit

3 .Adyen Payments Processor Tech company w/ banking license International customers Omnichannel

4 .Example payment

5 .Example payment

6 .Example payment

7 .Example payment

8 .Example payment

9 .Adyen

10 . Decision Points Update account? Do I know him? Weird behaviour? Strong authentication? How to retry? When to retry? Fraud? How to validate card?

11 .Data-Driven Products and Services Update account? Do I know him? Weird behaviour? Tokenization RevenueProtect ShopperDNA RevenueAccelerate Monitoring Strong authentication? How to retry? When to retry? Fraud? How to validate card? * few examples of many web services

12 . Live Platform Tokenization RevenueProtect ShopperDNA RevenueAccelerate Monitoring

13 . Live Platform Tokenization RevenueProtect ShopperDNA RevenueAccelerate Monitoring Highly Available Low Latency Sensitive Data

14 . Enter Data Platform Live Platform Data Platform

15 .Live Platform <> Data Platform Products 💸 ETL ML in production Magic ✨

16 .Live Platform <> Data Platform Products 💸 ETL ML in production Magic ✨

17 . ETL Pipelines Analytics Stream Filebeat Financial Stream Avro/Parquet Avro Derived tables

18 .Live Platform <> Data Platform Products 💸 ETL ML in production Magic ✨

19 . ML in Production BeamPush Alfred Push configurations Serve models

20 . ML in Production: Alfred Unified Framework • Build models • Track experiments Alfred • Re-train Serve models • Rollout • Serve • …

21 .ML in Production: Alfred Score Score Score Serve, Rollout Build/Experiment (Re)-training

22 . ML in Production: Alfred Rollout Performance Tracking (PT) PT PT Registered Ghost Test Live Retired Failed Fallback Model available Score and log, Requests to Requests to Replace by Trash, failed AT, in live platform Async new model, new model, newer model Not to be used Specified % 100%

23 .Scalable Time Series Forecasting and Monitoring using Apache Spark and ElasticSearch at Adyen Today 17h20 G102

24 .We are hiring! www.adyen.com/careers

25 .DON’T FORGET TO RATE AND REVIEW THE SESSIONS SEARCH SPARK + AI SUMMIT

6点赞

2收藏

0下载