On-Device ML Performance Engineer: SoC & Inference Expert

Apple, Inc.; Seattle (WA); On-site; USD 142,000 - 263,000; Full time

Job summary

Apple, Inc. in Seattle, WA, is seeking an On-device ML Performance Engineer to optimize ML model inference on Apple Silicon and the on-device stack across CPU, GPU, and Neural Engine.

You will work with PyTorch, MLX, and other frameworks, develop tools, analyze latency, memory, and power, and collaborate with research, software, and hardware teams to push performance boundaries.

Strong Python/C++ skills, experience with ML architectures, and system-level debugging are essential.

Qualifications

• Experience with ML inference, quantization, performance and accuracy.
• Familiarity with popular ML architectures (LLMs, diffusion, CNNs).
• Passion for latest advances in ML model design and on-device inference.
• Familiarity with Linux or macOS and embedded systems.
• Strong Python/C++ and shell scripting skills.

Responsibilities

• Drive on-device performance analysis of Apple SoCs and ML software stack across models.
• Optimize model conversion, compilation and on-device inference for energy efficiency.
• Develop tools and scripts to generate and analyze ML performance data.
• Collaborate across teams to deliver best-in-class on-device ML hardware and software stack.
• Present ML performance data to internal and external teams.

Skills

ML inference

Quantization

Performance analysis

Python

C++

Shell scripting

Linux

Education

Masters in CS

PhD in CS

Tools

CoreML

MPS Graph

MLX

PyTorch

ONNX

ExecuTorch

MLIR

LLVM

TVM

Xcode

Job description

Apple, Inc. in Seattle, WA, is seeking an On-device ML Performance Engineer to optimize ML model inference on Apple Silicon and the on-device stack across CPU, GPU, and Neural Engine.

You will work with PyTorch, MLX, and other frameworks, develop tools, analyze latency, memory, and power, and collaborate with research, software, and hardware teams to push performance boundaries.

Strong Python/C++ skills, experience with ML architectures, and system-level debugging are essential.