Michalis Papadimitriou from the TornadoVM team in UNIMAN presented at Devoxx Greece 2026 the latest advancements in running large language models (LLMs) on GPUs directly from Java. TornadoVM, as part of AERO, continues to drive open-source innovation by enabling high-performance execution of Java applications on heterogeneous hardware.
This presentation demonstrated how developers can build and run LLMs locally using GPUs through Java, leveraging TornadoVM and GPULlama3.java. It also provided insights into the programming model and practical steps for deploying AI workloads efficiently.
