Systems Software | High-Performance Computing | JVM Internals
Systems Engineer focused on high-performance computing, off-heap memory management, and hardware acceleration. Bridging modern JVM infrastructure (Project Panama, Vector API) with native execution environments (Rust, WebGPU) to architect zero-copy, high-throughput systems.
Engineered a high-performance FFI bridge connecting the JVM to WebGPU, enabling Java applications to execute GPU compute shaders without legacy C++ driver dependencies.
- Architecture: Designed a three-tier execution pipeline mapping Java off-heap
MemorySegmentallocations directly to a custom Rust/wgpu middleware layer via JDK Panama (FFM). - Performance Focus: Optimized for low-latency heterogeneous compute by enforcing zero-copy memory boundaries and direct GPU buffer mapping across platforms.
Developed a high-throughput numerical computing and tensor library designed to bypass legacy JNI bottlenecks using modern JVM systems features.
- Architecture: Leveraged JDK Panama (FFM) for contiguous off-heap memory allocation and multi-dimensional array management, ensuring strict cache-line alignment.
- Hardware Acceleration: Implemented CPU-level register optimizations and SIMD vectorization via the Vector API, utilizing AVX-accelerated matrix multiplications alongside parallel Fork/Join execution.
- Systems Java: JDK 22+, Project Panama (FFM), Vector API (SIMD), Project Valhalla, Project Loom
- Native & Middleware: Rust, C-ABI, Foreign Function Interface (FFI)
- Compute & Hardware: WebGPU,
wgpu, SIMD pipelines, Off-heap memory alignment - Testing & Benchmarking: JMH (Java Microbenchmark Harness), JUnit, Assembly instruction inspection