Principal Compute Architect

Morr0; San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Morr0 is seeking a Principal AI Compute Architect to lead the design of a new GPU/AI software stack, spanning compilers, runtimes and GPU execution strategies. The role focuses on optimizing kernels, SIMT execution and memory scheduling while collaborating with silicon and software teams on co-design.

The ideal candidate will have deep expertise in GPU compilers, CUDA, and AI runtimes, with a track record of influencing code generation and performance at scale in production environments.

Qualifications

• Significant experience in GPU architecture, compilers or ML systems.
• Strong C/C++ expertise and understanding of CUDA/SIMT.
• Experience with LLVM/MLIR, GPU kernel optimization or AI runtimes.
• Solid performance profiling, debugging and architectural decision-making skills.
• Ability to work across multiple layers of the AI compute stack.

Responsibilities

• Shape the architecture of a new GPU/AI software stack.
• Drive performance optimization across compilers, runtimes, kernels and hardware.
• Influence LLVM/MLIR, code generation and GPU execution strategy.
• Optimize scheduling, memory utilization and SIMT execution.
• Work closely with silicon and software teams on hardware/software co-design.
• Help scale the platform for production AI inference.

Skills

GPU architecture

C/C++ expertise

CUDA/SIMT

LLVM/MLIR

Performance profiling

Job description

Principal AI Compute Architect – GPU Software & Systems

We're partnering with a well-funded AI compute startup developing a next-generation GPU architecture and software platform.

We're looking for a senior architect with deep expertise in GPU compilers, runtimes, CUDA/kernel optimisation and AI systems, capable of understanding how AI workloads move from frameworks through to hardware execution.

What you'll do

• Shape the architecture of a new GPU/AI software stack.


• Drive performance optimization across compilers, runtimes, kernels and hardware.


• Influence LLVM/MLIR, code generation and GPU execution strategy.


• Optimize scheduling, memory utilization and SIMT execution.


• Work closely with silicon and software teams on hardware/software co-design.


• Help scale the platform for production AI inference.

What we're looking for

• Significant experience in GPU architecture, compilers, ML systems or low-level software.


• Strong C/C++ expertise and understanding of CUDA/SIMT.


• Experience with LLVM/MLIR, GPU kernel optimization or AI runtimes.


• Strong performance profiling, debugging and architectural decision-making skills.


• Ability to work across multiple layers of the AI compute stack.

Relevant technologies: PyTorch, JAX, LLVM, MLIR, CUDA, Triton, TensorRT, vLLM, CUTLASS, GPU ISA and AI accelerators.

You don't need experience across every layer. We're particularly interested in people with exceptional depth in GPU compilers, code generation, low-level GPU execution or hardware/software co-design who can think beyond their immediate specialism.

This is a Principal level individual contributor opportunity with significant ownership over the direction of a new AI compute platform.