11 months ago
Remote, United States +3 moreMid Level
Responsibilities
- Architect the migration of the existing compiler flow into MLIR, including dialects, passes, and lowering strategies.
- Build conversion paths between MLIR and Mythic’s custom low-level IR.
- Define MLIR validation infrastructure, including interpretation or execution paths for simulation and debugging.
- Extend MLIR integration across analog accelerators and digital subsystems.
- Leverage Torch-MLIR for PyTorch inputs and guide future integration with TorchDynamo and TorchInductor.
- Collaborate with hardware engineers, ML developers, runtime engineers, and compiler designers.
Requirements
- At least 3 years of experience in compiler or high-performance systems development.
- Proficiency in modern C++ standards including C++14, C++17, or C++20, and Python.
- Hands-on experience with MLIR, including dialect design, compiler passes, or lowering pipelines.
- Strong understanding of compiler intermediate representations and transformations, including lowering high-level operations to hardware-aware representations.
- Preferred experience architecting complete MLIR flows from frontend dialects through hardware-aware dialects and conversion to and from existing IRs.
- Preferred familiarity with PyTorch compiler technologies, especially Torch-MLIR, TorchDynamo, and TorchInductor.
- Preferred knowledge of dataflow architectures, scheduling, and memory orchestration.
- Preferred background with heterogeneous or specialized accelerators such as analog compute, NPUs, GPUs, or DSPs.
Benefits
- Opportunity to shape the MLIR and PyTorch integration strategy for novel AI hardware.
- Opportunity to contribute upstream and influence the broader ML compiler ecosystem.
- Collaboration with teams spanning hardware, runtime, and compiler design.
- Competitive compensation, equity, and comprehensive benefits.
