Compiler Engineer

OHO Group (Lithuania); San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

OHO Group in San Francisco is building an accelerator platform and the software to run demanding AI workloads. We are hiring compiler engineers who want to work close to the hardware, with ownership over how models become efficient executable code.

You will develop compiler passes, map computation to accelerator hardware, and collaborate with kernel and architecture teams to push performance. Experience with PyTorch, CUDA or GPU compilers is a plus.

Qualifications

• Strong C++ and intermediate representations experience.
• Experience with GPU/accelerator compilers and graph compilation.
• Familiarity with PyTorch, Triton, CUDA, runtimes or accelerators.

Responsibilities

• Develop compiler passes, lowering pipelines and code generation for AI workloads.
• Improve computation and memory access mapping onto accelerator hardware.
• Work across compiler/runtime interfaces, model enablement and performance debugging.
• Collaborate with kernel and architecture engineers to improve execution efficiency.
• Build tests and benchmarks that validate correctness and measurable performance gains.

Skills

C++

IR optimization

Code generation

GPU compilers

Graph compilation

Hardware optimization

PyTorch

CUDA

Triton

LLVM

Tools

LLVM

Job description

A well-funded AI compute startup is building a new accelerator platform and the software needed to run demanding AI workloads on it.

The team is hiring compiler engineers who want to work close to the hardware, with ownership over how models become efficient executable code. You’ll help bring workloads onto a new architecture, investigate performance bottlenecks and make decisions that influence both software and silicon.

What you’ll work on

• Develop compiler passes, lowering pipelines and code generation for AI workloads.
• Improve how computation and memory access map onto accelerator hardware.
• Work across compiler/runtime interfaces, model enablement and performance debugging.
• Collaborate with kernel and architecture engineers to improve execution efficiency.
• Build tests and benchmarks that validate correctness and measurable performance gains.

What we’re looking for

• Strong C++ and a grounding in intermediate representations, optimization and code generation.
• Experience with GPU or accelerator compilers, graph compilation, or hardware-aware optimization.
• Ideally, exposure to data-centre-class products and workloads.

Experience with PyTorch, Triton, CUDA, runtimes or custom accelerators is particularly useful.