The MTIA (Meta Training & Inference Accelerator) Software team has been developing a comprehensive AI Compiler strategy and optimizing compiler toolchains. This enables training and inference of Meta’s production DL/ML workloads on the specialized MTIA AI accelerator hardware in a highly performant and flexible way.
We are looking for a Software Engineering Manager who drives the compiler stack development & high performance compilers optimizations and tuning, specific to the MTIA AI accelerator hardware.
Software Engineering Manager, AI Compiler Responsibilities
- Grow a team of domain experts within AI Compiler.
- Communicate, collaborate, and build relationships with clients and peer teams to facilitate cross-functional projects.
- Operate strategically and tactically. Develop vision, strategy and help set direction for the team.
- Remain up-to-date on ongoing software development activities in the team, help work through technical challenges, and be involved in design decisions.
Minimum Qualifications
- Experience with compiler architecture and development, particularly ML compilers or DSLs or static/dynamic languages compilers.
- 2+ years of experience in managing a team of compiler engineers of varied skill levels.
- Experience with cross functional collaboration with hardware or AI framework teams.
- Demonstrated experience recruiting, building, structuring, leading technical organizations, including performance management.
Preferred Qualifications
- Experience with compiler optimizations such as loop optimizations, vectorization, parallelization, HW architecture specific optimizations.
- Experience in compiling and code generation targeting ML accelerators or custom hardware, GPUs or CPUs.
- Experience with different programming models for high-performance computations, e.g. GPU CUDA programming or OpenCL or OpenMP programming.
- Experience with MLIR, or LLVM or IREE or XLA or Triton or TVM or Halide.
- Knowledge of ML frameworks like PyTorch, TensorFlow, ONNX, MXNet, etc.