Meta
Compiler Engineer, Graph Compiler Performance Optimization
$184K–$257K
Bellevue, WA +2 · Full-time · Posted Aug 10, 2026
- Location
- Bellevue, WA
Menlo Park, CA
New York, NY - Experience
- Bachelor's degree: 7+ years; Master's degree: 4+ years; PhD: 3+ years of relevant experience in compiler development or accelerating deep learning models on hardware architectures.
- Education
- Bachelor's degree in Computer Science, Computer Engineering, or a relevant technical field, or equivalent practical experience; alternatively, a Master's degree or PhD.
About the job
Develop and optimize MTIA graph compiler passes that improve performance for AI workloads, collaborating with researchers and hardware and compiler teams.
What you'll do
- Design, implement, and validate graph-level compiler optimizations for the PyTorch Inductor / FX IR pipeline.
- Analyze model performance and develop fusion, memory, and tooling improvements to reduce latency and increase hardware utilization.
- Collaborate with hardware teams and AI researchers on hardware-software co-design and compiler support for new chip capabilities.
What they ask for
- Experience with C/C++ and Python programming.
- Experience in compiler development, performance optimization, or accelerating deep learning models on hardware architectures.
- Experience with graph-level compiler optimizations and performance profiling and analysis.
- Experience with AI hardware accelerators and deep learning frameworks.
Nice to have
- Experience with PyTorch internals and compilation stack (TorchDynamo, FX IR, Inductor), or similar frameworks such as XLA, TVM, MLIR, or Glow.
- Experience implementing responsible AI practices and integrating AI tools to improve workflows.
Pay
$183,997/year to $257,000/year USD base salary
Tools
C/C++, Python, PyTorch, TorchDynamo, FX IR, Inductor, XLA, TVM
AI focus
Optimize graph compilation and performance for AI workloads on MTIA hardware; support emerging model architectures and collaborate on model and compiler co-design.
Model-written summary · the original posting is authoritative