Current 2024 keynote
Jay Kreps · Sep 17, 2024 · 1h 37m · Austin, Texas
Summary
Kreps argued that AI applications need fresh, well-defined data, and that streaming "data products" published by each team are how companies get there. Flink was positioned as the processing layer for AI use cases, with Flink coming to Confluent Platform on premises and Tableflow exposing Kafka topics as Apache Iceberg tables with one click.
Key themes: Data platforms, Agents, Open source · AI-written summary from the published transcript
- 1AI applications need data that is relevant and up to date with what is happening right now, Kreps argued9:54
- 2Confluent has been adding direct model calls to Flink so streams can feed AI in real time15:08
- 3Flink was pitched as a strong platform for AI from simple chatbots to autonomous agents18:37
- 4Each team should publish its data as real-time streams, "data products" the rest of the company can reuse24:32
- 5Flink comes to Confluent Platform so the same processing runs in customers' own data centers1:12:20
- 6Tableflow exposes a Kafka topic as an Apache Iceberg table with one click1:04:04
Announced
| Announcement | Status |
|---|---|
| Flink in Confluent PlatformStream processing on premises | Announced |
| TableflowKafka topics as Apache Iceberg tables | Announced |
| Flink Table API for Java and PythonProgrammatic Flink in Confluent Cloud | Announced |
Related keynotes
| Date | Keynote | |||
|---|---|---|---|---|
| Oct 29, 2025 | Data platforms, Agents | Jay Kreps opened Current 2025 in New Orleans on building intelligent systems with real-time data | 8 |
About this keynote
Keynote page: www.youtube.com/watch?v=Sn6fVsOzrSU
- Jay KrepsCo-founder and CEO, Confluent
Companies named: Confluent
Last changed 2026-09-24.
Sources: official keynote videos, their captions and the hosts' recaps, summarized by AI; watch the source for exact wording. Logos via logo.dev; trademarks belong to their owners.