Staff AI Application Engineer - Heterogeneous Computing
New
A
Axelera AIAI computing
Work fully remotely from any European country (incl. the UK) you are already in.Full-TimeStaff
Salary not disclosed
Apply NowOpens the employer's application page
Job Details
- Experience
- 8+ years of engineering experience in systems, compilers, runtimes or performance
- Required Skills
- PythonC++
Requirements
- Have 8+ years of engineering experience in systems, compilers, runtimes, or performance.
- Have owned system-level design.
- Have hands-on heterogeneous computing experience across CPU, GPU, DSP, NPU, or FPGA.
- Be able to reason about compute, memory bandwidth, latency, and data movement.
- Have solid fundamentals in computer architecture, operating systems, and parallel programming.
- Be familiar with ML graph representations and compiler concepts, including graph capture, lowering, operator fusion, partitioning, and memory planning.
- Have strong C++ and Python skills.
- Use AI coding agents daily on real engineering work and be able to review and test their output.
- Preferred: have hands-on experience with MLIR or another compiler framework such as TVM, XLA, TorchInductor, or IREE.
- Preferred: have experience with an inference runtime or execution engine, GPU compute, or SIMD optimization on Arm or x86.
- Preferred: have experience with large-model execution techniques, agent environments, evaluation harnesses, verification frameworks, or MCP-based tooling.
Responsibilities
- Build reusable tools and environments for partitioning, compiling, running, and profiling models, including access to real hardware.
- Develop closed-loop model bring-up runs that apply changes, measure results against a reference, and repeat.
- Integrate differential testing and regression detection into CI.
- Own how PyTorch and ONNX models are captured, analyzed, and split across AIPU units, host CPUs, and host accelerators.
- Build readable and overridable placement tags in collaboration with the graph compiler team.
- Extend heterogeneous execution to large models, cascades, and multi-model pipelines, including expert offloading and caching.
- Work with application and runtime teams to improve end-to-end performance through overlap, zero-copy buffers, batching, and multi-core scheduling.
- Profile device time, host time, and data movement, and confirm speedups in the full pipeline.
- Study important customer models on Metis and Europa and turn findings into platform improvements.
View Full Description & ApplyYou'll be redirected to the employer's site
