Staff AI Application Engineer - Heterogeneous Computing - Hybrid/Remote - Europe (incl. UK), United Kingdom
Originally posted 26 October 2026 by the employer.
About the role
As a Staff AI Application Engineer in the Model Enablement team (Application Software), you will build the model enablement platform and determine how models run across the Metis or Europa AIPU, the host CPU, and host accelerators. The role spans architecture and hands-on execution across the model frontend, graph compiler, runtime, and application pipeline.
What you'll do
- Build the model enablement platform and agent harness to partition, compile, run, profile, and verify models with closed-loop regression testing in CI. - Own PyTorch and ONNX model capture, analysis, and partitioning across AIPU matrix and RISC-V vector units, host CPU, and host accelerators. - Extend heterogeneous execution architectures to large models, cascades, and multi-model pipelines using expert offloading and host/device memory caching. - Optimize end-to-end performance alongside application and runtime teams using zero-copy buffers, work overlap, batching, and multi-core scheduling. - Profile device execution, host processing, and data movement to resolve bottlenecks and validate pipeline speedups on Metis and Europa hardware.
What you'll need
- 8+ years of engineering experience in systems, compilers, runtimes, or performance, including system-level design ownership. - Hands-on experience with heterogeneous computing across CPU, GPU, DSP, NPU, or FPGA architectures, including compute, memory bandwidth, latency, and data movement analysis. - Solid fundamentals in computer architecture, operating systems, and parallel programming. - Familiarity with ML graph representations and compiler concepts, including graph capture (torch.export, FX, ONNX), lowering, operator fusion, partitioning, and memory planning. - Strong proficiency in C++ and Python. - Daily experience utilizing AI coding agents on production engineering work, with the ability to review and verify generated output.
Nice to have
- Hands-on experience with MLIR or other compiler frameworks such as TVM, XLA, TorchInductor, or IREE. - Experience with inference runtimes or execution engines, such as ONNX Runtime execution providers or custom NPU/DSP runtimes. - Experience with GPU compute (CUDA, OpenCL, Vulkan) or SIMD optimization on Arm or x86 architectures. - Experience with large-model execution techniques, including offloading, expert caching, speculative decoding, and inter-device pipelining. - Experience building agent environments, evaluation harnesses, verification frameworks, or MCP-based tooling.
Skills: graph compiler, runtime, heterogeneous computing, PyTorch, MLIR
This role has been open 0 days — well below the 67-day median for Infrastructure/Platform Software roles.
Infrastructure/Platform Software · AI ML Hardware
|
Open roles in category
106
|
Median days open
67 d
|
Median salary
$216k
|
See the full market breakdown ▾Category comparison, skills in demand, and who else is hiring
| Metric | Axelera AI | All employers we track in this specialty (106 roles · 28 employers) |
|---|---|---|
| Open roles in this specialty | 2 | 106 |
| Open roles in the wider Software, Firmware & Systems family | 5 | 6292 · 146 employers |
| Median days open | 0 d | 67 d (−67 d vs this employer) |
| Median salary (USD postings) | — | $216k |
Skills observed across this category: graph compiler, runtime, heterogeneous computing, PyTorch, MLIR
Who's hiring in this category
- Qualcomm · 25 open roles · median 57 d
- NVIDIA · 18 open roles · median 63 d
- Graphcore · 8 open roles · median 123 d
- AMD · 6 open roles · median 46 d
- Intel Corporation · 6 open roles · median 18 d
- Micron Technology · 6 open roles · median 50 d
How we counted: 106 open Infrastructure/Platform Software (AI ML Hardware) roles from 28 employers tracked in the SemiconductorJobs index, counted 27 Oct 2026. Specialty figures count only roles carrying this exact specialty label, so an employer's related work in neighbouring specialties is not included there — it is counted in the wider Software, Firmware & Systems family row. Figures refresh nightly.