Senior AI Kernel & Performance Engineer

Meta Careers; Menlo Park (CA); On-site; USD 154,000 - 217,000; Full time

Benefits offered by this job

Bonus

Equity

Benefits

Job summary

Meta is hiring an experienced kernel and performance engineer to own MTIA accelerator workloads, from the innermost loop of fused attention to numeric decisions that influence model convergence.

You will read hardware specs, implement high-performance kernels, and build kernel authoring frameworks and libraries so engineers can reach high performance across PyTorch operators, in eager and compiled paths.

Qualifications

• Proficiency in C++ and Python for low-level systems programming.
• Experience writing and optimizing kernels for parallel AI architectures.
• Strong understanding of memory hierarchies and performance measurement.

Responsibilities

• Design and optimize high-performance kernels for MTIA accelerators.
• Profile and root-cause performance across the stack from kernel to hardware.
• Develop kernel authoring frameworks and libraries for broad PyTorch operator coverage.
• Collaborate with silicon teams on hardware/software co-design.
• Bring up software on simulators, emulators, and early silicon.

Skills

C++

Python

Kernel optimization

Performance profiling

Parallel architectures

Tools

CUDA

ROCm

MLIR

LLVM

TVM

XLA

CUTLASS

cuDNN

Job description

Meta is hiring an experienced kernel and performance engineer to own MTIA accelerator workloads, from the innermost loop of fused attention to numeric decisions that influence model convergence.

You will read hardware specs, implement high-performance kernels, and build kernel authoring frameworks and libraries so engineers can reach high performance across PyTorch operators, in eager and compiled paths.