Quarkus

Quarkus is a Java framework designed for containers and Kubernetes, with an extension system for adding capabilities, published AI blueprints and guidance on using Java for AI workloads, working groups, a roadmap and documentation in English, Portuguese, Spanish, Chinese and Japanese.

Optimising for container density is the design goal and it addresses an economic problem rather than a technical preference. Java applications have historically used a great deal of memory at rest, which is irrelevant on a dedicated server and expensive in Kubernetes where memory requests determine how many containers fit on a node. Reducing the resting footprint changes the cost of running the same application, which is an argument a platform team makes to a finance team.

The extension system is what makes that achievable. Extensions perform work at build time that a traditional framework does at startup, so adding a capability does not add startup cost, and the ahead-of-time compilation path becomes viable.

Publishing AI blueprints reflects where the framework's sponsors expect Java to be used next, and it is a reasonable position: a great deal of enterprise data sits in Java systems, and moving the model to the data is often easier than the reverse.