Soirée TornadoVM [en]

Date et lieu

Les sessions sont filmées et le public est photographié. Les photographies sont ensuite publiées sur le site du Paris JUG et autres médias de l’associations Paris JUG. En acceptant cette invitation, vous autorisez le Paris JUG à publier votre photo sur les médias sus-mentionnés.

Les inscrits à l’évènement non présents 5 minutes avant le début de la session, soit à 19h25, verront leurs places remises à disposition.
Les non inscrits à l’évènement ne pourront donc y assister que sous réserve de places disponibles sur place 5 min avant le début de la session, soit à 19h25.
L’inscription implique de posséder une adresse mail valide sur laquelle vous recevrez une demande de confirmation à laquelle il vous faudra répondre afin de valider votre inscription. Toute inscription non confirmée ne sera pas prise en compte !

Inscription

Détails

19h00 : Accueil

19h30 : Write Once, Accelerate Everywhere: GPU-Ready Java with TornadoVM

In the AI era, accessing GPUs from Java is essential - but it still remains a challenge. TornadoVM bridges this gap by enabling developers to accelerate Java applications on GPUs, CPUs, and FPGAs without writing any GPU code. Originally born as a PhD research project at the University of Manchester, TornadoVM has evolved into a mature open-source framework supported by both academia and a growing global community.

This talk will present the key innovations that are offered by the TornadoVM runtime platform, and will focus on practical ways to optimize existing Java applications through its high-level API. Through hands-on examples, we will demonstrate how to offload computations, manage data efficiently, and achieve significant performance gains - entirely from Java.

We will also showcase jitLLM, a pure Java large language model (LLM) library accelerated with TornadoVM, illustrating how modern AI workloads can run efficiently within the JVM. Finally, a live-coding demo using IntelliJ IDEA will highlight TornadoVM’s integrated profiling and tooling ecosystem for seamless development experience.

Thanos Stratikopoulos

Dr. Thanos Stratikopoulos is a Principal Software Engineer at Beestream Ltd and Research Fellow at the University of Manchester with specialization on heterogeneous architectures and reconfigurable accelerators. He has authored more than 20 research articles in the field of hardware acceleration, system software and programming languages. He has also been involved in several EU Horizon 2020, EU Horizon Europe, and UKRI projects and his interests include computer architecture, high-performance computing, virtualization and cloud acceleration.

20h30 : Buffet offert par Criteo

21h00 : Building LLM Inference Engines in Java with TornadoVM and the NVIDIA Ecosystem

As Java enters the AI era, developers are looking for ways to run local large language models (LLMs) without leaving the JVM. This session presents lessons learned from building GPU-accelerated LLM inference engines natively in Java using TornadoVM, an open-source framework that transparently offloads Java programs to GPUs and other accelerators.

Beyond generating optimized CUDA code directly from Java, TornadoVM now provides seamless access to the NVIDIA software ecosystem, enabling Java applications to leverage libraries such as cuBLAS, cuDNN, cuFFT, and CUDA runtime APIs without requiring JNI or manually written native code. This allows developers to combine high-level Java programming with the performance of NVIDIA’s AI and HPC libraries.

Using jitLLM, a lightweight GPU-accelerated inference engine for Llama models, the session explores techniques for supporting quantized data types (FP16, Q8, and Q4), integrating with Quarkus and LangChain4j, and optimizing LLM inference directly in Java. Along the way, attendees will see how modern Java can target GPUs while interoperating with the broader NVIDIA ecosystem. The session concludes by discussing how TornadoVM complements emerging JVM technologies such as Project Babylon and Project Panama, and how these efforts are helping establish Java as a first- class platform for AI, accelerated computing, and high-performance machine learning.

Thanos Stratikopoulos

Dr. Thanos Stratikopoulos is a Principal Software Engineer at Beestream Ltd and Research Fellow at the University of Manchester with specialization on heterogeneous architectures and reconfigurable accelerators. He has authored more than 20 research articles in the field of hardware acceleration, system software and programming languages. He has also been involved in several EU Horizon 2020, EU Horizon Europe, and UKRI projects and his interests include computer architecture, high-performance computing, virtualization and cloud acceleration.

22h00 : 3ème mi-temps des juggers

Feedback

Sponsors de l'évènement

Platinium

Code de Conduite

Les conférences Paris JUG se veulent des expériences sans harcèlement, quel que soit votre sexe, votre identité sexuelle, votre âge, votre orientation sexuelle, votre handicap, votre apparence physique, votre poids, votre race ou votre religion. Nous ne tolérons aucun harcèlement des participants à la conférence, quel que soit sa forme. Les expressions et les images à connotation sexuelle ne sont pas appropriées lors des événements ou sur les médias en ligne. Les participants à une conférence qui violent ces règles peuvent être sanctionnés, voire exclus de la conférence.