Principal Software Architect

Oho Group; San Francisco (CA); On-site; USD 140,000 - 190,000; Full time

Job summary

OHO Group (Lithuania) is seeking a GPU Software Architect to drive performance-focused GPU software across compiler, kernel, and runtime interfaces. You will profile workloads, identify bottlenecks in execution and memory behavior, and optimize workload mapping and memory access.

The role emphasizes hands-on implementation, collaboration with hardware and software specialists, and responsibilities spanning compiler lowering, code generation, and kernel optimization.

Qualifications

• 3+ years of GPU, accelerator or systems software experience.
• Strong C++ and CUDA or comparable GPU programming experience.
• Practical understanding of GPU execution and memory hierarchies.
• Depth in compiler development or performance-critical GPU kernels.
• A measured approach to debugging and optimisation.

Responsibilities

• Develop and optimise GPU software across compiler, kernel and runtime interfaces.
• Profile workloads and investigate bottlenecks in execution and memory behaviour.
• Improve workload mapping, occupancy, synchronisation and memory access.
• Contribute to compiler lowering and code generation or kernel implementation and optimisation.
• Work with hardware and software specialists on design, debugging and validation.

Skills

GPU programming

Kernel/Compiler development

Performance optimization

C++

Tools

CUDA

Job description

A team building advanced compute technology is hiring a GPU Software Architect to work across performance-critical software and its interaction with hardware.

Building on the rapid progress they have made, our client is is looking for people with experience working across one or more layers of the software stack, with an emphasis on kernels or compilers.

This role suits engineers who want more technical ownership while staying close to implementation.

What you’ll do:

• Develop and optimise GPU software across compiler, kernel and runtime interfaces.
• Profile workloads and investigate bottlenecks in execution and memory behaviour.
• Improve workload mapping, occupancy, synchronisation and memory access.
• Contribute to compiler lowering and code generation or kernel implementation and optimisation.
• Work with hardware and software specialists on design, debugging and validation.

What you’ll bring:

• 3+ years of relevant GPU, accelerator or systems software experience.
• Strong C++ and CUDA or comparable GPU programming experience.
• Practical understanding of GPU execution and memory hierarchies.
• Depth in compiler development or performance-critical GPU kernels.
• A measured approach to debugging and optimisation.