I run Java on GPUs — building GPU-accelerated runtimes, enabling the Nvidia Ecosystem on the JVM, compilers, and LLM inference engines for the JVM.
- Position: Senior Software Engineer @ Neo4j (Cypher Runtime) · Research Fellow @ The University of Manchester
- Interests: JIT Compilation, GPUs, Managed Runtimes, Machine Learning Compilers, LLM Inference, Database Performance
- Software Stack: Java, C++, Scala, Python, CUDA, OpenCL, Metal, GraalVM, Apache TVM, TornadoVM
- jitLLM — lead author. GPU-accelerated LLM inference (Llama3, Mistral, Qwen, Phi-3, Granite) in pure Java via TornadoVM. OpenCL / CUDA / Metal backends, LangChain4j + Quarkus integration.
- TornadoVM — core maintainer. A Java framework for transparently offloading JVM applications to GPUs, FPGAs, and multi-core CPUs without rewriting them in CUDA or OpenCL.
- Neo4j — contributor to the Cypher Runtime, working on query execution and performance-critical database runtime components.