Adding AI to an existing Java application doesn’t have to mean rebuilding the entire system around a model. With the right boundaries, AI features can become another well-defined part of the application. Complete with tests, observability, deployment, and rollback.

Brian Benz, Principal AI Advocate at Microsoft, brings 13+ years of experience across AI, cloud, and open-source technologies, while Ed Burns, Principal Architect at Microsoft, has worked on enterprise Java technologies since the 1990s and led major specifications including Servlet and JavaServer Faces. Together, they show how to add document search and a guarded AI assistant to an existing Spring Boot/Jakarta application without losing the engineering practices teams already rely on.

You’ll leave the session with a practical approach to integrating AI into enterprise Java using structured outputs, evaluation tests, OpenTelemetry, containers and Kubernetes while treating model failures, secrets, version changes, and rollback as normal application concerns.

Getting an application into production is only half the job. Once it’s running, developers need to understand what is happening inside it. Especially when performance drops, requests fail, or services start behaving differently under load.

With more than 20 years of experience in Java development, architecture, DevOps, cost optimization & observability, Thiago Gonzaga has spent years helping teams build scalable systems and understand their behavior in production. As an Oracle ACE Associate and leader of the Java Noroeste User Group, he combines deep Java experience with hands-on cloud-native practices.

You’ll leave this session with practical examples for instrumenting Quarkus applications with OpenTelemetry, collecting traces, metrics, and logs, and connecting them to an LGTM stack with Loki, Grafana, Tempo and Mimir. Giving you a clearer path from “it works on my machine” to actually understanding what happens in production.

Your CI pipeline signs releases, holds credentials, and runs third-party code. Yet many teams still protect it far less carefully than production. Attackers know that, which is why trusted tooling and build infrastructure have become such attractive targets.

Brian Demers is a Java Champion, Apache Member and Developer Advocate at Gradle with deep experience in build automation, open source and software security. In this session, he maps the trust boundaries of both CI pipelines and AI-assisted developer environments, using real incidents like Codecov and tj-actions to show where risk actually enters the build.

You’ll leave with a practical way to assess your own build pipeline and clear hardening priorities: pin versions, scope tokens, use ephemeral runners, and keep build records that help you investigate what happened later.

Adding AI to an enterprise application doesn’t have to mean introducing a new language, a separate stack, or rewriting what already works. With Jakarta EE, MicroProfile, and langchain4j-cdi, Java developers can bring AI capabilities directly into the technologies and patterns they already use.

Emily Jiang, Java Champion, long-time MicroProfile & Jakarta EE specification leader and Ivar Grimstad, Java Champion, Jakarta EE Developer Advocate and EE4J PMC Lead, both help shape the standards behind modern enterprise Java. Together, they show how AI can be integrated directly into CDI-managed applications without leaving the Java ecosystem.

You’ll leave the session with practical techniques for augmenting APIs, automating business logic, and building AI-powered services on top of your existing Java infrastructure without adding Python or rebuilding your application from scratch.

Regional outages shouldn’t automatically mean downtime. For modern applications, active-active architectures can remove the dependency on a single primary region and make resilience part of the system design instead of an emergency procedure.

As a Spring Developer Advocate at Broadcom, DaShaun Carter focuses on helping developers build, run and manage better software with Spring. In this session, he takes a single-region Spring Boot application and turns it into a geo-distributed deployment using active-active In-Memory Data Grids and a global load balancer.

You’ll leave with a practical understanding of why cross-region data calls become a problem, how active-active architectures improve resilience, and how to design a Spring Boot application that can keep running even when an entire region disappears.

High-performance AI in Java has traditionally meant JNI bindings, native wrappers, and a lot of integration complexity. Modern Java APIs such as the Foreign Function & Memory API and Vector API are changing that by giving Java applications more direct access to GPUs and SIMD-optimized CPU execution.

As the creator of Neuroph and Deep Netts, a Java Champion, Duke’s Choice Award winner and member of the JCP Executive Committee, Zoran Sevarac has spent years bringing AI capabilities directly into the Java ecosystem.

You’ll leave this session with a practical look at how IBM Semeru Runtime with OpenJ9, NVIDIA CUDA, FFM, and the Vector API can work together to build Java-native AI applications that combine GPU acceleration with optimized CPU inference without relying on JNI.

AI can make individual developers faster, but the bigger challenge is turning that advantage into something an entire team can use consistently. Without shared practices, context, and engineering discipline, AI adoption easily stops at isolated experiments and vibe coding.

With more than 20 years of software engineering experience, Luiz Real combines hands-on work with Java and AI with years of teaching, mentoring, and community leadership as a SouJava JUG leader. His focus on software engineering fundamentals gives him a practical perspective on how teams can adopt AI without losing the practices that make software maintainable and reliable.

You’ll leave the session with a clearer approach to context engineering, AI agents, and harness engineering and practical ideas for turning individual AI usage into a repeatable team capability that helps software delivery scale.

As AI agents take on more implementation work, the developer’s role starts to shift from writing every line of code to expressing intent, shaping context and reviewing what agents produce. That changes what we should expect from an IDE.

Bruno Borges, Principal Product Manager at Microsoft, works at the intersection of software development, AI, cloud and product strategy. His experience guiding large-scale developer products gives him a strong perspective on how development environments need to evolve as agentic workflows become part of everyday engineering.

You’ll leave the session with a practical look at intent-driven development, including executable specifications, parallel agent workflows, isolated worktrees, model selection, context management, automation and human approval boundaries and a clearer idea of what the IDE of the agentic age may look like.

Open source infrastructure feels free because developers rarely see the cost of every download, build and cache miss. At scale, though, small inefficiencies can create enormous load. Especially when CI pipelines, ephemeral environments and tooling defaults repeatedly fetch the same artifacts.

Brian Fox, co-founder and CTO of Sonatype, former Chair of the Apache Maven project, Apache Software Foundation member and OpenSSF Governing Board member, has spent more than 20 years helping shape the infrastructure behind modern software development. Using Maven Central as a real-world case study, he shows how usage patterns can quietly push shared infrastructure to its limits.

You’ll leave the session with a better understanding of where unnecessary load comes from, how build and cloud practices contribute to it and what teams can change to use open-source infrastructure more responsibly without giving up speed or reliability.

Agentic AI systems don’t just generate answers, they can take actions, call tools, access data and make decisions. That creates a different security model, where mistakes can have consequences far beyond a bad response.

Micah Silverman, Director of AI Acceleration at Snyk, combines 30 years of Java experience with a long career in application security. His background across software engineering, security and developer education gives him the perspective to translate emerging AI risks into concrete engineering concerns.

You’ll leave the session with a practical understanding of the OWASP Top 10 for Agentic Applications (2026), including real-world examples of each risk and the defensive controls you can use to design safer agentic systems from the start.