13:05 | 13:45
Keywords defining the session:
- Stream processing
Takeaway points of the session:
- Stream processing is necessary for modern fraud detection and marketing. To author and monitor streaming processes you don't have to be developer
- Nussknacker and Flink are enough for analysts.
Analyzing and gaining insights from large amounts of data is one thing. Doing it in real time is a whole different business.
There quite a few advanced stream processing engines, Apache Flink is one the most widely used. However, designing, testing and deploying streaming jobs usually demand development skills – it’s not so easy for analysts or business people.
When we introduced Flink in one the largest polish telcoms (which is used for real time marketing and fraud detection there) we decided to build open source project – Nussknacker – to make it easier for non-programmers to use all Flink power.
I’d like to tell about it and show what we’ve built. The talk will be based on real-world examples – from simple filterings to aggregations and model serving.