Exploring the breadth and depth of Adoptium technologies

We hit an important milestone in July by surpassing 1 billion downloads. Amazing, but not a signal for us to "rest on our laurels". This presentation showcases some key initiatives in our project and how we continue to innovate and build upon the technical achievements that help us keep pace.

Enterprise AI is booming, but relying on external LLM services introduces serious risks. Security and compliance policies often prohibit sending proprietary data outside the organization. At the same time, latency and fast-growing API costs, driven by the exploding token use of AI agents, make external dependence unsustainable for many businesses.

Running AI locally is a strong alternative. But how can we safely integrate local models into mission-critical enterprise systems on a trusted and secure Java runtime, while still reusing existing Java assets and meeting critical regulatory requirements for AI and data security?

This session presents a practical 3-tier Java architecture, built upon an Adoptium-based runtime, for integrating local models into mission-critical enterprise systems using Jakarta EE:

• Low-level execution: While Project Babylon is still evolving, we show how TornadoVM, already used in production, can accelerate AI workloads on GPUs. This architecture utilizes an Adoptium-based runtime, ensuring software supply chain security.
• Enterprise framework layer: Jakarta EE and MicroProfile expose AI capabilities with reliability, observability (metrics, tracing), and governance aligned with audits and regulatory compliance.
• AI integration layer: LangChain4j, together with LangChain4j-CDI for seamless Jakarta EE integration, and ONNX Runtime Java orchestrate local models, manage, and implement agentic workflows. We also show how these components can power a MCP server inside a Java stack.

Through a live demo, you will learn how to assemble these layers on a secure Adoptium-based runtime into a secure, low-latency AI architecture that keeps sensitive data on-premise while delivering modern LLM capabilities.

This presentation will showcase LogLens, an AI-powered log analysis assistant designed to automatically review Jenkins server logs, system logs (such as syslog), and other infrastructure outputs to identify warnings, errors, and anomalous patterns in real time. It utilizes NLP and ML techniques and a custom storage system utilizing DuckDB.

A short recorded team presentation done by the CMU students working for a term from June-Aug on Agentic Workflows designed to help committers and the PMC manage the project more effectively.

Providing developers with secure, consistent and high-quality tooling is essential to delivering software predictably across a global organization. As part of this mission, Accenture's global Java Community of Practice transformed the Java developer workstation to modernize the developer experience, improve software supply chain transparency and reduce operational complexity.

This session shares the journey of redesigning the Java workstation standard around Eclipse Adoptium's Temurin, Mission Control and other key Eclipse Foundation projects. Attendees will learn how the new workstation architecture was defined, validated and deployed, the approach used to transition from legacy Java distributions and the governance processes established to ensure enterprise-wide adoption. The presentation will also explore the business and engineering outcomes of the initiative, including improved asset provenance, simplified developer onboarding and maintenance, greater platform consistency and meaningful cost efficiencies achieved at global scale.

Through practical lessons learned and real-world implementation insights, this talk provides a blueprint for organizations seeking to modernize their Java development environments while strengthening security, governance and developer productivity.

Redesigning a production CI pipeline for one of the world's most widely downloaded JDK distributions is no small undertaking. In this lightning talk, Andrew Leonard shares how he is leveraging IBM Bob AI as a core engineering partner — across system design, architecture decisions, and hands-on code development — to drive the ground-up redesign of Adoptium's next-generation CI pipeline infrastructure.

The current Adoptium build pipelines have served the project well, but the demands of a growing, multi-platform, multi-version OpenJDK distribution require something better architected: modular, maintainable, and CI-agnostic. Andrew will walk through how IBM Bob AI has accelerated this rearchitecture effort — from early exploratory design conversations and structural decision-making through to generating, reviewing, and iterating on real implementation code — and share honest observations on where AI assistance genuinely shifts the pace and quality of complex infrastructure engineering.

Attendees will leave understanding:

• What a CI-agnostic, modular build pipeline architecture looks like for a project at Adoptium's scale
• How IBM Bob AI can be practically applied across the full engineering lifecycle — not just code generation
• The real-world experience of using AI as a design and architecture partner on a large open-source infrastructure project

Artificial Intelligence is rapidly becoming part of every developer's toolkit, but how can it provide real value within a large-scale open source infrastructure project?

This session explores how the Adoptium Infrastructure Team is applying AI and data-driven automation to improve visibility, troubleshooting, maintenance and development workflows across the Temurin build and test infrastructure.

Rather than focusing on theory, this talk demonstrates practical tools and projects currently being developed, including infrastructure capacity monitoring, environment auditing, AI-assisted code reviews, and intelligent log analysis capable of producing guided recommendations for Jenkins failures.

Can you truly trust a binary you didn't build yourself? Eclipse Temurin answers that challenge head-on with its 3rd Party Reproducible Verification Build process — a supply chain trust mechanism that allows independent parties to rebuild Temurin JDK binaries from scratch and cryptographically verify they are bit-for-bit identical to the published artefacts.

In this lightning talk, Andrew Leonard walks through how Adoptium has extended its secure supply chain story beyond SBOM generation: from independent toolchain guidance (GCC DevKits, matching Visual Studio and Xcode versions) through CycloneDX SBOM-embedded build recipes, to signed CycloneDX Attestation XML files that record the evidence of a successful — or failed — verification. Andrew will show what verified reproducible builds look like in practice through the publicly visible results on Adoptium.net, giving attendees a tangible picture of supply chain transparency at scale.

Attendees will leave understanding:

• Why reproducible builds matter for enterprise JDK consumers
• How CycloneDX Attestations formalise third-party verification evidence
• How independent toolchain guidance enables willing third parties to participate in supply chain verification