The CFP is now open. Join us in San Francisco on October 7–8, 2026.
Something has shifted in the way we think about AI agents.
A year ago, the conversation was about prompts, chains, and demos. Today, it's about production. The teams building serious agent systems are running into a familiar set of problems: how do you give an agent real-time context? How do you persist memory across sessions? How do you trigger actions when something happens in the outside world? How do you orchestrate tool calls reliably, at scale, with full observability?
These look like new problems. In many ways they are. But the shape of them — durable state, event ordering, replayable history, multi-tenancy, observability at scale — is deeply familiar to anyone who has spent the last decade building streaming infrastructure.
That's why this year's Data Streaming Summit is expanding. AI agents demand real-time context and new infrastructure. A decade of building streaming systems is how we know what to build. The summit is now the meeting point for two communities working on the same layer of the stack from different directions: the data streaming community building the engines that move events at scale, and the agent infra community building the runtimes, harnesses, and observability layers that make autonomous systems production-ready.
Both groups are solving infrastructure problems. Increasingly, they're solving the same problems. This is where they meet.
What's New This Year
Data Streaming Summit has always been a gathering place for the streaming community — practitioners running Pulsar and Kafka at scale, engineers pushing the limits of stream processing, and the operators behind real production systems. Past summits have featured speakers from OpenAI, Netflix, LinkedIn, Uber, Confluent, Databricks, and many more.
This year, we're expanding the conversation to include the engineers, founders, and platform teams building the agent era. Not as a side track, but as a first-class part of the summit's identity.
The new positioning is reflected in three tracks, each representing a layer of the stack:
Data Streaming Engines — the foundation everything else runs on. Talks on the architecture, internals, and operations of streaming systems: Pulsar, Kafka, Ursa, Flink, Spark, RisingWave, and the next generation of primitives like diskless topics, lakehouse-native storage, and leaderless log protocols.
Data Streaming for Agents — the bridge layer. How streaming infrastructure powers production agent systems: real-time context, durable memory, event-driven tool orchestration, agent-to-agent messaging, and the observability pipelines that capture every trace, span, and decision.
Agent Harness, Runtime, and Governance — the application layer. The runtimes, orchestration frameworks, evaluation harnesses, and governance patterns that make agents reliable and safe in production. LangGraph, AgentCore, Codex, Claude Code, and the in-house runtimes enterprise teams are building today.
The three tracks read top-to-bottom as a single story: from the engines at the bottom, to the streaming substrate that connects them to agents, to the runtimes that make agent systems trustworthy at scale. This is the infrastructure layer of the AI era, and we're building it together.
Why This Matters
Most AI conferences today are about models. Most infrastructure conferences are about the systems of five years ago. There's a gap in the middle — the place where production AI meets production infrastructure — and that's where the most interesting work is happening.
We've been seeing this conversation in fragments for the past year. A team at one company solves the agent memory problem with an event log. A team at another company replaces their LLM observability pipeline with a streaming telemetry system. A platform team rebuilds their RAG architecture as a real-time pipeline because batch processing isn't fast enough for agents. The agent runtime team across the hall starts asking the streaming team how to do durable execution and exactly-once tool calls. These aren't isolated stories — they're the early signal of a category forming.
Data Streaming Summit 2026 is where these stories get told in one room.
The Call for Speakers Is Open
We're looking for talks that are technically deep, grounded in production, and willing to share real numbers. If you're running streaming infrastructure at scale, building agent systems in production, or working on the runtime and governance layers that make agents enterprise-ready, we want to hear from you.
You don't need to come from the streaming community to belong here. If you're building agent infra and finding yourself wrestling with state, ordering, replay, observability, or any of the problems streaming engineers have spent years on — this is your conference too.
We especially welcome submissions that bridge the two communities — case studies where streaming infrastructure made an agent system possible, or where agent workloads pushed streaming systems in new directions.
CFP closes June 30, 2026. Submit your talk here.
Speakers receive a free conference pass, full promotion across our channels, and a platform to engage with the practitioners building the infrastructure layer of the AI era.
Save the Date
Data Streaming Summit 2026 October 7–8, 2026 Hotel Nikko, San Francisco
Whether you're a longtime member of the Pulsar and Kafka communities, a platform engineer integrating streaming with AI workloads, or an agent builder trying to take your system from demo to production — this is your conference.
We'll see you in San Francisco.
Want to sponsor, exhibit, or partner on the summit? Reach out to the StreamNative events team at events@streamnative.io.
