Building GPU-Accelerated AI Libraries in Java with TornadoVM, Thanos Stratikopoulos

This talk discusses how Java is evolving in the AI landscape, focusing on the development of high-performance machine learning workloads within the JVM. The speaker shares insights from building GPU-accelerated AI libraries using TornadoVM, a framework that enables Java programs to utilize GPUs and other accelerators. By examining GPULlama3.java, a library designed for large language models, the session covers techniques for handling quantized data types such as FP16, Q8, and Q4, along with integration strategies for Quarkus and LangChain4J. Attendees will learn practical approaches to optimize AI inference in Java and how TornadoVM enhances Project Babylon and Project Panama, positioning Java as an advanced platform for AI development.