BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//sf26-agenda//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:SESS-144@sf26-agenda
DTSTAMP:20261020T151709Z
DTSTART:20261104T210000Z
DTEND:20261104T214500Z
SUMMARY:Using Agentic AI to create the next generation of non-Java Kafka cl
 ients
LOCATION:Room 2010
DESCRIPTION:Shivsundar R\, Andrew Schofield\n\nConfluent's non-Java Kafka c
 lients handle over 350 billion connections a month to Confluent Cloud. The
 y're built on librdkafka\, a C library that has served the community well 
 for over a decade. But every new KIP follows the same painful path: origin
 al implementation in Java\, wait months for someone to reimplement it in C
 \, then wrap it for Python\, .NET\, JavaScript and Go. Some features never
  arrive at all. Over 40% of Kafka users are stuck in this permanent featur
 e gap. And users encounter subtle behavioral and configuration differences
  between the Java and non-Java clients which just causes confusion. What i
 f we stopped reimplementing and started translating? What if AI-assisted p
 roductivity could enable us to attempt something we would have never attem
 pted previously? What if we could leverage AI tools to take the Java refer
 ence implementation and produce a working Rust client\, with bindings for 
 all five languages\, within weeks of a Java release? That's what we built.
  This talk is a deep dive into using AI to write production-grade systems 
 code -- not prototypes\, not demos -- but a Kafka client that has to handl
 e the same scale and correctness bar as the library it replaces. Why this 
 is hard. Keeping five clients in sync with Java isn't about hiring more en
 gineers. KIPs touch protocol handling\, serialization\, group coordination
 \, and configuration. Reimplementing each one across languages is slow and
  error-prone. The AI translation pipeline. How we built a system of AI age
 nts that listens for commits in the Apache Kafka Java code\, translates it
  to idiomatic Rust\, and generates language-specific wrappers. The careful
 ly crafted prompts\, meticulous context management\, agent orchestration\,
  and verification layer that catches when the model gets it wrong. What AI
  gets wrong in systems code. Handling situations where a direct translatio
 n is not good enough\, subtle behavioral divergences that only surface und
 er load\, and why you can’t blindly trust AI-generated code to be produc
 tion-ready. Come and hear about how we used AI and what we learnt as engin
 eers about how to get the best out of the AI tools.
URL:https://sf26.rmoff.net/s/using-agentic-ai-to-create-the-next-generation
 -of-non-java-144
END:VEVENT
END:VCALENDAR