Staff AI Application Engineer - Heterogeneous Computing
Core
Build the model enablement platform that decides how AI models run across heterogeneous hardware (AIPU, CPU, accelerators) to maximize end-to-end performance for the Voyager SDK.
Role type
Staff AI Application Engineer (Heterogeneous Computing)
Builds
Model enablement platform, agent harness, and execution pipelines for the Metis/Europa AIPU.
Domain
AI Infrastructure / Heterogeneous Computing / Systems Programming
Deliverable
production ML models
Required skills
System-level design, heterogeneous computing (CPU/GPU/DSP/NPU/FPGA), computer architecture, parallel programming, ML graph representations (PyTorch/ONNX), C++, Python, AI coding agents.
Preferred skills
MLIR/TVM/XLA/TorchInductor/IREE, inference runtimes, GPU compute (CUDA/OpenCL/Vulkan), large-model execution techniques, agent environments/verification frameworks.
Technologies
PyTorch, ONNX, MLIR, TVM, XLA, TorchInductor, IREE, CUDA, OpenCL, Vulkan, C++, Python.
Responsibilities
Build tools to partition, compile, run, and profile models; manage model ingestion and splitting across AIPU matrix/RISC-V/CPU/accelerators; extend execution for large models and multi-model pipelines; optimize end-to-end performance via scheduling and memory management; study critical customer models to improve the platform.
Seniority
Staff, hands-on IC with architecture ownership.
