
AERO is a @HorizonEU RIA project on open source software that will enable the deployment of cloud services on the outcomes of the @EuProcessor.
Joined January 2023
- Tweets1.2K
- Following226
- Followers432
- Likes1.4K
Pinned Tweet
📣 New Press Release: AXHEL, an open-source C++ library from @DII_Pisa accelerating the arithmetic behind Homomorphic Encryption on AArch64.
👉 aero-project.eu/wp-content/u…
Repo: github.com/AERO-Project-EU/A…
@EU_HaDEA
@HorizonEU
@hipeac
@EU_CloudEdgeIoT
#opensource #he #encryption
jitLLM 3.0 is out! #Java #LLM inference on CUDA, Metal and OpenCL GPUs via #TornadoVM.
• 1.43× faster startup with @graalvm Native Image
• Up to 9.3× faster prefill
• Zero-copy weights on @Apple silicon
Soon in @langchain4j @QuarkusIO @springcentral
github.com/beehive-lab/jitll…
AERO retweeted
Next Tuesday (17 Nov) I'm at @parisjug for a TornadoVM evening!
☕ GPU-ready Java, no GPU code needed
🤖 LLM inference in pure Java with jitLLM
🟩 CUDA + NVIDIA libs from Java, no JNI
19:00 at Criteo. Register 👇
parisjug.org/events/2026/11-…
#Java #TornadoVM #GPU #LLM
AERO retweeted
mlxVS: cuVS for your Mac. GPU vector search on Apple silicon, built on MLX. 1M vectors indexed in 10 s on a MacBook Pro, then 70k queries/s at 97.6% recall, 2x compared to CPU HNSW. github.com/mikepapadim/mlxvs
AERO retweeted
Our next #YouTube video shows 4 ways to install #TornadoVM.
SDKMAN! with 1 command, prebuilt archives for Linux, macOS and Windows, building from source, & #JBang with no SDK at all.
👉 youtu.be/_ipmOKOYj_s
Subscribe: youtube.com/@TornadoVM-Dev
#Java #GPU #TornadoVM #tutorial
AERO retweeted
We started a #YouTube channel for #TornadoVM.
Subscribe here: youtube.com/@TornadoVM-Dev
First video: how to take a plain Java app and run its hot methods on the GPU, in four steps: find, transform, compose, execute.
👉youtu.be/BjiEJvNusIk
#opensource #Java #GPU #tutorial
AERO retweeted
The goal behind the new feature of #TornadoInsight shipped in 2.1.0, is to use LLMs to transform plain #Java methods into #GPU-accelerated @tornadovm tasks, all without leaving @intellijidea. And every kernel is verified against the original method before you accept it.
TornadoInsight 2.1.0 is out 🌪️
New (experimental): #AI turns a plain #Java method into a #TornadoVM kernel, then runs it on your #GPU and checks it against the original.
Black-Scholes on an Apple M4 Pro: 25× faster than Java.
#IntelliJ @intellijidea
plugins.jetbrains.com/plugin…
AERO retweeted
TornadoInsight 2.1.0 is out 🌪️
New (experimental): #AI turns a plain #Java method into a #TornadoVM kernel, then runs it on your #GPU and checks it against the original.
Black-Scholes on an Apple M4 Pro: 25× faster than Java.
#IntelliJ @intellijidea
plugins.jetbrains.com/plugin…
AERO retweeted
New @TornadoVM release, v8.0.0 is coming with support for ahead-of-time compilation via @graalvm Native Image and Project Leyden.
tornadovm.org/blogs/fast-sta…
Huge leap for the project! Kudos to all contributors! 🙌
#opensource #Java #GPUs #TornadoVM
TornadoVM 8.0.0 is out 🌪️
After 10 years: thank you, Graal. Java kernels are now compiled by TC, our own pure-Java compiler to CUDA, OpenCL and Metal.
• AOT ready
• No JVMCI
• Now Apache 2.0
• Your code doesn't change
👉 tornadovm.org/blogs/tornadov…
#OpenSource #Java #AI #GPU
AERO retweeted
TornadoVM now runs Groovy code on GPUs. Great to see wider adaption of the tornadovm as the go-to-way to access GPUs from the JVM!
Groovy code running on the GPU with TornadoVM: Metal and OpenCL on an Apple M4 Pro, CUDA on an NVIDIA T4. A 3840×3200 Mandelbrot in ~12 ms vs ~2.6 s on 12 CPU cores.
groovy.apache.org/blog/groov… #groovylang #TornadoVM #GPU
@ApacheGroovy @tornadovm @TheASF
AERO retweeted
TornadoVM 8.0.0 is out 🌪️
After 10 years: thank you, Graal. Java kernels are now compiled by TC, our own pure-Java compiler to CUDA, OpenCL and Metal.
• AOT ready
• No JVMCI
• Now Apache 2.0
• Your code doesn't change
👉 tornadovm.org/blogs/tornadov…
#OpenSource #Java #AI #GPU
AERO retweeted
Had so much fun at @JDD_Krakow 2026 in Kraków yesterday.
I presented “Java Goes CUDA-Native”: new @tornadovm features that let @jitLLM run LLM inference at GPU-native speed, from Java.
tornadovm.org
jitllm.io
#opensource #Java #TornadoVM #GPU #AI
AERO retweeted
LLM Inference on GPU in Java?
@tornadovm and @jitllm at @Devoxx Belgium!
* Java writing Java that runs on the GPU. Java for the GPU, on the GPU!.
* A @QuarkusIO @langchain4j application reviewing it's future port to Quarkus 4 locally, all in Java on GPU!
youtu.be/HO5CpETzywk?is=_ws2…
AERO retweeted
Yesterday my colleagues @MXekalaki and @mikepapadim showed live demos about using @nvidia cuFFT and cuBLAS with @tornadovm for signal filtering and @jitllm!
NVIDIA performance is a jar away from Java.
#Java #CUDA #TornadoVM
Optimized GPU libraries (e.g. cuBLAS, cuDNN, etc.) by @nvidia, now one jar away for your Java - without caring about JNI.
We also show demos with @ApacheFlink and @jitllm.
Watch our talk in @Devoxx 👇
youtu.be/6SPnBmQrYEk?si=3DI_…
#opensource #AI #TornadoVM #Java #GPU
AERO retweeted
Optimized GPU libraries (e.g. cuBLAS, cuDNN, etc.) by @nvidia, now one jar away for your Java - without caring about JNI.
We also show demos with @ApacheFlink and @jitllm.
Watch our talk in @Devoxx 👇
youtu.be/6SPnBmQrYEk?si=3DI_…
#opensource #AI #TornadoVM #Java #GPU
jitLLM is part of the @langchain4j 1.21.0 release! 🚀
Read more here: docs.langchain4j.dev/integra…
#opensource #LLM #GPU #Java #LangChain4j #Inference
#LangChain4j 1.21.0 is out! 🚀
Meet Decision Models: instant yes/no, choice and scale answers with probabilities.
Decision Services turn any Java interface into a classifier.
Plus model routing, re-ranking, tool selection, guardrails and agentic routing!
github.com/langchain4j/langc…
AERO retweeted
Next week we will be in @Devoxx and speak about @tornadovm, @jitllm and how we accelerate Java on GPUs, using @nvidia native libraries!
Come and talk to us, to learn more!
#opensource #Java #AI #TornadoVM #jitLLM #GPU
AERO retweeted
TornadoVM has a new family member! #jitLLM provides blazing fast AI inference at native speed⚡️👇
Can Java compete with llama.cpp/@ollama? YES! With @jitLLM!
jitLLM uses @nvidia CUDA libraries (cuBLAS, cuDNN) and reaches 0.90× to 1.12× of llama.cpp's throughput!
jitLLM is empowered by @tornadovm and will be soon in @langchain4j and @QuarkusIO.
github.com/beehive-lab/jitll…
AERO retweeted
Excited to see our work reaching this point. ⚡
Pure Java LLM inference, seriously fast. Approaching - and in some cases exceeding - llama.cpp performance.
Can Java compete with llama.cpp/@ollama? YES! With @jitLLM!
jitLLM uses @nvidia CUDA libraries (cuBLAS, cuDNN) and reaches 0.90× to 1.12× of llama.cpp's throughput!
jitLLM is empowered by @tornadovm and will be soon in @langchain4j and @QuarkusIO.
github.com/beehive-lab/jitll…
























