MTIA Performance Auto-Tuning Lead

Meta; Menlo Park (CA); On-site; USD 210,000 - 320,000; Full time

Job summary

Meta is seeking an experienced engineer to lead performance autotuning on MTIA, Meta's custom training and inference accelerator. You will head the MTIA Software Performance Autotuning team within Infra Foundations and own how we extract maximum performance from hardware, across kernels, graphs, and runtime configurations.

You will set autotuning strategy, design the search and benchmarking framework, and collaborate with compiler, kernel, and product teams to ship speedups.

Qualifications

• Bachelor's degree in CS/CE or equivalent practical experience.
• Strong Python and C++ skills with hands-on experience across the PyTorch stack.
• 4+ years in ML systems, AI infra, performance engineering, or similar.

Responsibilities

• Lead performance autotuning on MTIA to maximize hardware and software efficiency.
• Define autotuning strategy, build search and benchmarking infrastructure, and guide cross-team collaboration.
• Partner with compiler, kernel, runtime, and product teams to ship speedups.

Skills

Python

C++

ML systems

Cross-team collaboration

Mentoring

Performance engineering

Roofline analysis

Experimentation

Technical leadership

Education

BS in CS/CE or related field

MS/PhD in CS/CE or related field

Tools

TorchInductor

torch.compile

MLIR

XLA

TVM

Triton

Ansor

AutoTVM

Job description

Meta is seeking an experienced engineer to lead performance autotuning on MTIA, Meta's custom training and inference accelerator. You will head the MTIA Software Performance Autotuning team within Infra Foundations and own how we extract maximum performance from hardware, across kernels, graphs, and runtime configurations.

You will set autotuning strategy, design the search and benchmarking framework, and collaborate with compiler, kernel, and product teams to ship speedups.