On-device ML Compiler Engineer, Model Compilation, Graphics, Games & ML

Apple Apple · Big Tech · Cupertino, CA · Machine Learning and AI

This role focuses on building and optimizing ML compilers and runtimes for on-device execution across Apple's diverse hardware (Neural Engine, GPU, CPU). It involves working with MLIR-based compiler stacks to improve runtime performance and enable efficient execution of ML models on Apple devices, impacting core experiences like Camera, Siri, and Health.

What you'd actually do

  1. Inspire changes in our MLIR-based compiler in order to target improved runtime performance by demonstrating the capabilities of the hardware.
  2. Propose upstream changes in MLIR to better support new features and workflows in the hardware that lead to more optimal execution performance across all types of devices and device clusters.
  3. Own core pieces of the compiler stack enabling heterogeneous compute across Apple devices. We target execution of ML models across the Apple ecosystem from resource-constrained devices like Apple Watch, to the high-end Macs with Ultra SoCs.
  4. Work closely with hardware, software, and performance teams across the company to accelerate and optimize execution by taking advantage of the latest features in the hardware, OS, and drivers.

Skills

Required

  • MLIR-based compilers
  • common ML model architectures
  • common ML model execution schemes
  • common ML model operations
  • C++
  • PyTorch or related training frameworks

Nice to have

  • Swift
  • programming paradigms for the GPU
  • programming paradigms for the CPU
  • programming paradigms for the Neural Engine
  • writing kernels for ML model execution

What the JD emphasized

  • MLIR-based compilers
  • runtime performance
  • ML models
  • execution

Other signals

  • ML infrastructure
  • ML compilers
  • model optimization
  • on-device ML
  • Apple silicon