GPU Software Engineer

GPU Software Engineer

Campbell, CA | On-site

Experience: 5+ Years

About OXMIQ

OXMIQ provides complete hardware and software GPU IP, so our customers can build their own AI silicon. Our OxCore architecture combines scalar, tensor and SIMT engines on one platform. It offers CUDA compatibility and stays flexible and fully programmable. Our software stack, OxPython, runs on top of it and lets CUDA-optimized models run unmodified on Oxmiq’s GPU IP, which gives day-zero support for new AI models. We have an established software team and a working stack. We’re now moving from development to production delivery: hardening the platform and scaling it to meet real customer demands.

You’ll join the OXMIQ software team as a hands-on engineer working across the GPU software stack. You’ll build, optimize and deliver the kernels, runtime and enablement layers that bring high performance to our customers’ AI workloads. We hire at multiple levels - Final job title, internal leveling, and compensation will be determined based on the candidate's total relevant experience, technical expertise, and interview performance.

Key Responsibilities

• Design, develop and optimize GPU software across OxPython, the OXMIQ software stack, covering kernels, runtime, driver and framework integration layers.
• Develop and tune high-performance GPU kernels (CUDA, Triton-based and custom ops) for AI workloads, including GEMM, attention, convolution and memory-bound operators.
• Drive performance optimization: profile workloads, find bottlenecks across kernels, runtime and driver, and close the gap to peak hardware utilization.
• Implement and optimize SIMT execution and parallel compute: thread and warp scheduling, divergence handling, occupancy and efficient use of the memory hierarchy.
• Contribute to work orchestration across OxCore’s scalar, tensor and SIMT engines, and to enabling CUDA-optimized models to run unmodified on Oxmiq hardware.
• Enable and bring up new accelerator features, models and customer workloads on Oxmiq GPU IP, from pre-silicon and emulation through production.
• Work with compiler engineers on MLIR/LLVM lowering and code generation as it affects kernel and runtime performance.
• Work with architecture and hardware teams to turn performance requirements into efficient software, and report software findings back to inform hardware definition.
• Support software/hardware validation, including test development, debugging and co-verification of the stack and silicon.
• Build benchmarking, profiling and root-cause analysis tools and practices for kernels and runtime-generated code.
• Harden software for production by focusing on reliability, scalability and performance across customer workloads, deployment sizes and model complexity.
• Take part in design and code reviews, and help mentor engineers across the kernels, runtime and frameworks teams.

Required Qualifications

• 5+ years building systems software for GPUs, accelerators or other heterogeneous compute hardware.
• Strong hands-on experience with GPU programming models, including CUDA, and with writing and optimizing high-performance kernels.
• Deep understanding of SIMT and parallel compute architectures: thread/warp execution models, divergence, synchronization, occupancy constraints and memory hierarchies.
• Strong expertise in C++ and systems programming, with experience in GPU runtimes, drivers or accelerator enablement.
• Proven track record in performance optimization: profiling, bottleneck analysis and tuning for compute- and memory-bound workloads.
• Hands-on experience integrating with ML frameworks (PyTorch, JAX, ONNX) and an understanding of how deep learning workloads are expressed and executed.
• Working knowledge of compiler infrastructure (LLVM, MLIR) and code generation concepts as they relate to kernel performance.
• Strong debugging skills across the software/hardware boundary and a bias toward shipping: you prototype, measure and iterate.
• Hands-on experience with Claude Code or equivalent AI-assisted development workflows.

Preferred Qualifications

• Experience with GPU software or runtimes for custom AI accelerators, NPUs or GPU IP.
• Experience with Triton or similar high-level GPU kernel frameworks, and with kernel fusion, operator tiling and auto-tuning.
• Experience delivering CUDA compatibility or porting CUDA workloads to non-NVIDIA hardware.
• Exposure to quantization, mixed-precision or sparsity-aware kernel and runtime techniques.
• Experience with performance modeling and roofline analysis for GPU workloads.
• Background in AI workloads and software stacks: LLM inference and training, attention and KV-cache optimization, distributed and multi-GPU execution, or serving frameworks such as vLLM and SGLang.
• Experience bringing up software on new or pre-silicon hardware (simulators, emulators, FPGA) and scaling it into production.
• Contributions to open-source GPU, compiler, runtime or ML-framework projects such as LLVM, MLIR, Triton or PyTorch.

Education

BS/MS/PhD in Computer Science, Computer Engineering, Electrical Engineering or a related field.

Why OXMIQ

You’ll join a fast-growing company with an established team and a proven stack. We’re scaling that stack into a production platform used by customers, and you’ll work directly with the people designing the hardware it runs on. This role is for an engineer who wants to work hands-on across an entire SIMT compute platform, not just one layer of it, and who measures success in shipped, high-performance software.

Compensation & Benefits

OXMIQ offers a competitive compensation package, including base salary, equity participation, comprehensive medical, dental and vision coverage, and the opportunity to contribute to foundational silicon and software technology.

OXMIQ is an equal opportunity employer. We evaluate qualified applicants without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, veteran status, or any other legally protected characteristic.