Senior AI Compute Architect — Full-Stack, High-Performance

OXMIQ Labs, Inc.; Campbell (CA); On-site; USD 180,000 - 280,000; Full time

Job summary

OXMIQ Labs, Inc. is hiring a Full-Stack AI Compute Architect to lead the software and silicon stack from models to kernels. You will influence OxPython, graph capture, compilers, and runtime, ensuring day-zero CUDA-compatibility with high-performance goals.

You will shape MLIR/LLVM backends, architecture of OxCore engines, and end-to-end optimization for large-scale inference across modern AI workloads.

Qualifications

• 12+ years building systems and AI infrastructure software with architecture ownership.
• Breadth across AI software stack from model level to kernel/instruction level.
• Hands-on experience with PyTorch, JAX, ONNX and deep learning models.
• Strong expertise in C++ and compiler infrastructure, including LLVM and MLIR.
• Deep SIMT/parallel compute expertise and GPU programming experience.

Responsibilities

• Refine the OxPython software stack direction in collaboration with the team.
• Map AI models to the platform and optimize capture, lowering, and inference.
• Drive end-to-end performance across graph, compiler, runtime, and kernels.
• Design graph/compiler strategy, MLIR dialects, and LLVM backends.
• Architect work orchestration across OxCore SIMD engines and CUDA-compatibility.
• Advance SIMT execution, scheduling, memory hierarchy, and occupancy.
• Optimize kernel programming models, tiling, and fusion with kernel engineers.
• Scale solutions for production and real customer workloads.

Skills

C++

LLVM

MLIR

CUDA

SIMT

GPU architectures

PyTorch

JAX

ONNX

Education

BS/MS/PhD in Computer Science/Computer Engineering/Electrical Engineering

Tools

LLVM

MLIR

CUDA

Triton

Job description

OXMIQ Labs, Inc. is hiring a Full-Stack AI Compute Architect to lead the software and silicon stack from models to kernels. You will influence OxPython, graph capture, compilers, and runtime, ensuring day-zero CUDA-compatibility with high-performance goals.

You will shape MLIR/LLVM backends, architecture of OxCore engines, and end-to-end optimization for large-scale inference across modern AI workloads.