Senior AI Kernel & Performance Engineer

Meta; New York (NY); On-site; USD 184,000 - 257,000; Full time

Benefits offered by this job

Equity

Bonus

Benefits

Job summary

Meta designs and deploys its own AI systems, MTIA—the Meta Training and Inference Accelerator—across Meta's data centers. The MTIA Software team operates within the AI & Compute Foundation, building the stack from compiler toolchains to PyTorch integration and leading kernel-level performance work.

You will own performance for workloads that serve billions, reading RTL-adjacent docs and ensuring the hardware is the bottleneck to be fixed, not the software.

Qualifications

• Bachelor's degree in Computer Science, Computer Engineering, or equivalent practical experience.
• 6+ years of professional experience in high-performance computing, accelerator kernel development, compiler backends, or systems performance engineering.
• Proficiency in C++ and Python, including low-level systems programming and performance-critical code.
• Experience writing kernels for parallel architectures (GPU, TPU, AI ASICs, or SIMD).
• Knowledge of memory hierarchies, bandwidth, and performance measurement.
• Ability to build roofline or analytical models and explain gaps.

Responsibilities

• Design, implement, and optimize high-performance compute kernels for MTIA accelerators.
• Profile and root-cause performance across the full stack from instruction scheduling to memory behavior.
• Extend kernel authoring frameworks and libraries for broad usage.
• Mentor engineers on performance methodology and accelerator programming.
• Collaborate with silicon teams on hardware/software co-design and graph production wins.
• Work with compiler, runtime, and product teams to deliver end-to-end improvements.
• Investigate numerics and precision trade-offs and design mitigations.
• Set technical direction for kernel domains and write design documents.

Skills

C++

Python

Performance analysis

GPU/ASIC experience

Templates/Generic programming

English communication

Education

Bachelor's degree in CS/CE or equivalent

Tools

CUTLASS

Triton

cuDNN

LLVM

MLIR

Job description

Meta designs and deploys its own AI systems, MTIA—the Meta Training and Inference Accelerator—across Meta's data centers. The MTIA Software team operates within the AI & Compute Foundation, building the stack from compiler toolchains to PyTorch integration and leading kernel-level performance work.

You will own performance for workloads that serve billions, reading RTL-adjacent docs and ensuring the hardware is the bottleneck to be fixed, not the software.