AI Software Stack Data Flow Engineer

About OXMIQ

OXMIQ designs GPU and AI silicon for large-scale model inference and training. The associated software stack connects industry-standard frameworks — PyTorch, JAX, ONNX, and Triton — to OXMIQ hardware and is central to the performance the platform delivers to customers.

The Role

The AI Software Stack Data Flow Engineer is responsible for the end-to-end flow of an AI workload through the OXMIQ software stack, from the program a customer writes to its execution on silicon. The role spans the major authoring entry points exposed by the modern AI software stack — frameworks, kernel DSLs, and low-level kernels — and is responsible for ensuring that each lands cleanly on OXMIQ hardware.

The role is hands-on. The engineer traces data flow across stage boundaries, identifies points at which it stalls or fails, and ships the code required to make the workload run efficiently.

The use of AI-assisted development tools (Claude Code or equivalent) is a required part of the role and is expected to be applied throughout the engineer's daily work, including code authoring, review, debugging, and documentation.

What Data Flow Means in This Role

A workload reaches OXMIQ hardware through multiple authoring paths — framework-level programs, kernel DSLs, and low-level kernels. The engineer is responsible for how each of these paths is captured, compiled, dispatched, and executed on the silicon. The work spans the graph compiler, the kernel-authoring layer, and the runtime, and the engineer is expected to identify and address performance and correctness issues at whichever stage of the data flow matters most for a given workload.

Key Responsibilities

- Own end-to-end data flow for AI workloads from source program through execution on OXMIQ hardware.

- Work across the framework, kernel-DSL, and low-level kernel authoring paths to ensure each lands cleanly on the silicon.

- Contribute to the graph compiler and kernel-authoring layers of the software stack.

- Drive performance through profiling, benchmarking, and targeted fixes at the appropriate layer of the stack.

- Use Claude Code or equivalent AI-assisted workflows as a first-class part of engineering practice.

- Collaborate with compiler, kernel, runtime, and hardware architecture teams.

Required Qualifications

- 5+ years of professional software engineering experience in the AI software stack.

- Working experience with PyTorch.

- Working experience with Triton or a comparable Python-based GPU kernel DSL.

- Working experience with CUDA, including the ability to read, debug, modify, and write kernel-level code.

- Strong C++ skills (modern C++17/20) and strong Python skills.

- Working understanding of the role of a graph compiler in the modern AI stack.

- Solid grounding in code generation and kernel performance fundamentals.

- Strong debugging and performance-analysis skills.

- Demonstrated proficiency with Claude Code or equivalent AI-assisted development workflows in day-to-day engineering practice.

- Comfort in a Linux development environment with standard build and CI tooling.

Preferred Qualifications

- Experience integrating a backend into a major ML framework (PyTorch, JAX, or comparable).

- Experience with MLIR or LLVM at a contributor level.

- Experience with vendor kernel libraries at the integration or authoring level.

- Experience shipping for a custom AI accelerator, NPU, or GPU IP target.

- Familiarity with quantization, mixed-precision, and sparsity-aware compilation and kernels.

- Experience with performance modeling and autotuning frameworks.

- Open-source contributions to relevant compiler, kernel, or framework projects.

Education

- BS/MS/PhD in Computer Science, Computer Engineering, Electrical Engineering, or a related field — or equivalent practical experience.

Working Environment

The engineer will work within a compact, senior team comprising compiler engineers, kernel authors, runtime engineers, and hardware architects, and will operate across the boundaries of those teams as the data-flow problem requires. AI-assisted development tools (Claude Code or equivalent) are a standard part of engineering practice at OXMIQ and are expected to be used in daily work.

Compensation & Benefits

OXMIQ offers a competitive compensation package, including base salary, equity participation, comprehensive medical 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.