Apache Kafka Sydney

Hello everyone! Join us for an Apache Kafka® meetup on Oct 13th from 5:30pm, hosted by Mantel in Sydney!

The address, agenda, and speaker information can be found below. See you there!

Venue:
Mantel
Level 21, 580 George St, Sydney NSW 2000

***
Agenda:

• 5:30pm: Doors open
• 5:30pm - 6:00pm: Pizza, Drinks, and Networking
• 6:00pm - 6:30pm: Jaehyeon Kim, Senior Data Engineer, Kasada
• 6:30pm - 7:00pm: Chris Arthur, Customer Success Engineering Manager, Confluent
• 7:00pm - 7:30pm: Additional Networking

***
Speaker:
Jaehyeon Kim, Senior Data Engineer, Kasada

Talk:
From Prototype to Production: Building a Real-Time ML Feedback Loop with Kafka and Flink

Abstract:
While many platforms power real-time decisions, most machine learning workflows remain stubbornly batch-oriented. This session presents a practical case study on redesigning an offline Python-based recommendation prototype into a highly scalable, real-time contextual bandit system built entirely on Apache Kafka and Apache Flink.
We will explore how Kafka serves as the backbone of the feedback loop by managing real-time event ingestion and user feedback streams. From there, we will dive into how Apache Flink performs continuous, stateful online model training directly within the stream. Join this session to learn how to safely manage ML model state inside a distributed stream processor, decouple continuous Flink training from low-latency serving, and successfully integrate online learning workloads into your event-driven architectures.

-----
Speaker:
Chris Arthur, Customer Success Engineering Manager, Confluent

Talk:
Atomic Writes Across Kafka and Your Database: What KIP-939 Changes

Abstract:
If you've ever built a system that writes to both Kafka and a database, you've hit the dual-write problem: how do you guarantee both writes succeed or fail together? Today's answer is usually the outbox pattern — a reliable but heavyweight workaround involving extra tables, pollers, and CDC pipelines.

In this talk, we'll cover what 2PC participation actually means for Kafka's transaction protocol, how it differs from Kafka's existing exactly-once semantics, and what this unlocks for teams building event-driven systems that need real atomicity between Kafka and external stores.