Front End Compiler

The role

You will own the ingestion layer of a high-performance, portable AI compiler. You will build systems to parse, validate, and lower representations from frameworks like PyTorch, StableHLO, ONNX, and MLIR dialects into an internal compiler IR.

What you will do

• Design, develop, and maintain the front end of the heterogeneous AI compiler
• Build and extend ingestion pipelines for ML frameworks and representations
• Define and evolve the interface between external model representations and internal compiler IR
• Ensure correctness and completeness of operator coverage
• Collaborate with graph optimization and code generation teams
• Use performance and correctness data to identify gaps in coverage
• Contribute to documentation and tooling for ML engineers

Required qualifications

• BS degree in Computer Science or equivalent practical experience
• 2+ years of experience working with ML optimization tools/libraries
• Experience working with or ingesting from ML frameworks such as PyTorch, TensorFlow/JAX, or ONNX
• 2+ years of experience with C/C++
• 2+ years working with Python
• Strong written and verbal communication skills
• Ability to write clear and concise technical documentation
• Knowledge of quantization, sparsity, or other model-level optimization techniques

Nice to have

• Master's or PhD in Computer Science or equivalent
• Experience with StableHLO, XLA, or other ML-specific IRs
• Knowledge of operator semantics across ML frameworks and the challenges of cross-framework compatibility
• Familiarity with Python bindings and tooling for compiler front ends
• Experience with testing infrastructure for compiler correctness — fuzzing, differential testing, or model-level validation
• Contributions to open-source compiler or ML framework projects
• Familiarity with MLIR, including defining and working with custom dialects

Benefits

• Equity
• Medical/dental/vision
• Retirement savings
• Wellness benefits
• Company bonus opportunities

About the company

Lemurian Labs is reimagining the foundations of computing to make AI accessible to everyone. They are building a hardware-agnostic platform to remove limits of scale, hardware, and cost.