Microsoft Reactor Tel Aviv

----------------------------------------------
Explore the repo

📢 Join the conversation on Discord!

📌 This session is a part of the Model Mondays series. Learn more here!

📌 This session is a part of a series! Learn more here

This session is a part of a series! Learn more here

## 🌎 Europe, Middle East & Africa Stream
Program order to come; please check jdconf.com for the latest details

---

1. Self-Improving Agentic Systems with Spring AI: From Context Curation to Autonomous Refinement

Speaker: Christian Tzolov (Broadcom)

AI agents are systems dedicated to the art of context window curation—relentless loops of assembling context, prompting the model, observing results, and re-assembling for the next step.

This talk explores how Spring AI's Advisors provide an elegant framework for building self-improving agentic systems. Advisors intercept, modify, and enhance AI interactions, encapsulating common patterns like memory management and RAG in reusable, portable components.

But traditional single-pass advisors cannot support the iterative workflows that real agents demand: tool calling loops, output validation with retry, and evaluation-driven refinement. Recursive Advisors solve this by enabling controlled iteration through the advisor chain.

Using Spring AI's fluent ChatClient API, we'll demonstrate practical examples including ToolCallAdvisor for explicit tool execution and LLM-as-a-Judge evaluators. You'll learn to build agents that don't just accomplish objectives, but iteratively improve how they accomplish them—achieving quality control through self-correction and feedback-driven refinement.

---

2. Modern Java Puzzlers

Speaker: Simon Ritter (Azul Systems)

Since JDK 9, Java has evolved faster than at any point in its 30-year history. For developers, this is great as it means more new features to make code more concise, yet hopefully easier to read and understand. All new language features also bring their quirks and possibly unseen consequences.

In this fun interactive session, we'll examine a range of recent language features and pose questions to the audience about how the code works and what it does. We'll cover many of the new language features from recent releases, including switch expressions, sealed classes, and pattern matching. We'll even look at some curious behavior in JShell and how some special characters can be used in unexpected ways.

The answers will most definitely not always be as you expect.

---

3. How the Dutch IRS Implemented App Modernization via AI-Assisted Development

Speaker: Ron Veen (Team Rockstars IT)

Working in one of the many teams at the Dutch IRS, we ran into a number of challenges while migrating our code base from a monolith application running on a mainframe to a set of microservices running on Kubernetes.

We started using AI-assisted coding to:

• Modernize our application
• Remove anti-patterns from our software
• Improve throughput
• Accelerate development of certain features
• Integrate modules better

Working in an organization with such sensitive information brought its own challenges, as our code base was not allowed to be uploaded or examined via regular web AI providers. Instead, we turned to Copilot Studio.

This talk discusses our experiences, challenges, pitfalls, and lessons learned during this migration, providing the audience with valuable insights into how to approach similar migrations.

---

4. Playwright: Your Next Java Test Framework for Automating Web Tests

Speaker: Alex Soto (IBM)

Playwright is a testing framework created by Microsoft to write tests for modern web applications. It helps create multi-browser/platform resilient tests that can be executed in different languages (Python, JavaScript, Java, and more), working in headless mode.

Apart from specific testing features, Playwright comes with a set of tools that enhance the developer experience: reports to inspect test execution, the state of the browser at each moment, why a test failed, and screen recordings of execution to visually inspect the test path. An automatic code generator is also provided to get started with Playwright smoothly.

---

5. You're Absolutely Right, It Was Your Home Directory!

Speaker: Oleg Šelajev (Docker)

Letting an AI agent loose for solving development tasks is a productivity dream—until it decides to optimize your home directory or brick your system by upgrading Python. We want that YOLO mode speed, but without the security nightmares.

In this session, we'll look at Docker Sandboxes: a new primitive designed to let agents operate in a restricted environment with limited access to the filesystem and controlled network and secret injections. We'll dive into the typical mess AI agents create, see why basic isolation isn't enough, and walk through a workflow for running agents that you can actually trust.

---

6. Java RAG Made Easy with Spring AI and Elasticsearch

Speaker: Laura Trotta (Elastic)

Retrieval Augmented Generation (RAG) can be a useful tool to turn a generic chatbot into a specialized expert in any desired subject, even with data that must remain private. Java isn't the first language people think of when tackling AI-related projects, and here's where Spring comes to the rescue with a new AI-dedicated library that leverages the power of vector stores such as Elasticsearch to provide an easy and intuitive experience for building AI applications. Expect just a bit of theory and a decent chunk of practice through a live demo.

---

7. Pipelines to Production: AI-Accelerated CI/CD for Java

Speaker: Brian Benz (Microsoft)

Delivery speed increases when AI assists with DevOps pipelines. This session shows CI/CD patterns where AI tools and generative AI assistants streamline Java builds: test generation, threat-aware code reviews, manifest creation and maintenance for Kubernetes, and auto-tuned deployments. It explains guardrails to prevent unsafe changes, demonstrates pipeline environment creation, and integrates observability for release validation. Attendees learn how to shorten feedback loops without compromising reliability or security.