Session Overview
How does South Korea’s largest telecom carrier process live activity across 30+ million mobile subscriptions to trigger marketing campaigns at the exact right moment?
In this session, data engineers from SK Telecom share how they built their real-time event processing and decisioning architecture. Learn how CAMA uses Apache Flink to filter 2 million events per second down to high-value triggers, and how AUF (AI Operating Framework) pairs real-time context with AI agents to automate campaign targeting and execution.
What You'll Learn
- Real-time stream filtering with Apache Flink: How CAMA standardizes raw logs and runs parallel scenario applications to match complex user behavior.
- Session gap calculation: How app usage time and engagement sessions are calculated on the fly from log timestamps.
- Architecture overview: Handling end-to-end data flow from on-prem data lakes and Apache Flink to AWS Kinesis, Kafka, and downstream messaging channels.
- AI agents in marketing operations: Transitioning from manual SQL campaign setups to AI-driven rule generation and autonomous campaign loops.
- Real-world results: How real-time contextual triggers lifted campaign CTR compared to traditional batch marketing.
Chapters
- 00:00 — Introduction & SK Telecom overview
- 02:41 — What is CAMA? Real-time marketing context
- 05:56 — Stream filtering at scale with Apache Flink
- 09:44 — Session gap & usage time calculations
- 11:07 — System architecture & data flow
- 12:33 — AUF: Automating campaigns with AI agents
- 19:19 — Results & autonomous marketing roadmap
- 21:31 — Q&A wrap-up
