Staff AI Platform Engineer — Heterogeneous Compute

Axelera AI; United States; Remote; USD 160,000 - 260,000; Full time

Benefits offered by this job

Pension plan

Comprehensive insurance

Stock options

Remote-friendly policies

Job summary

Axelera AI is seeking a Staff AI Application Engineer to build the model enablement platform and orchestrate model execution across Metis/Europa AIPUs and host accelerators. You will design frontend, graph compiler, runtime, and pipeline interfaces with a focus on performance and scalability.

You will drive model ingestion, planning, and verification, collaborating with graph compiler teams and device-side engineers to optimize end-to-end latency and throughput for diverse workloads.

Qualifications

• 8+ years in systems, compilers, runtimes or performance with design ownership.
• Hands-on heterogeneous computing across CPU, GPU, DSP, NPU or FPGA.
• Strong fundamentals in computer architecture, OS and parallel programming.
• Familiarity with ML graph representations and compiler concepts: graph capture, lowering, partitioning.

Responsibilities

• Model enablement platform and agent harness construction for model bring-up and verification.
• Model ingestion: capture, analyze and split models across AIPU units, host CPU and accelerators.
• Heterogeneous execution extension to large models and multi-model pipelines.
• End-to-end performance optimization with overlap, zero-copy, batching and scheduling.

Skills

System design ownership

Heterogeneous compute

Computer architecture

C++

Python

AI agents tooling

Tools

MLIR

TVM

XLA

CUDA

Job description

Axelera AI is seeking a Staff AI Application Engineer to build the model enablement platform and orchestrate model execution across Metis/Europa AIPUs and host accelerators. You will design frontend, graph compiler, runtime, and pipeline interfaces with a focus on performance and scalability.

You will drive model ingestion, planning, and verification, collaborating with graph compiler teams and device-side engineers to optimize end-to-end latency and throughput for diverse workloads.