Software Engineer, Kernel Performance & AI Tooling

Responsibilities

• Build developer tooling and workflows that make kernel development and performance optimization faster, more scalable, and easier to debug, integrate, and deploy.
• Develop observability, diagnostics, and validation infrastructure that makes AI-assisted optimization systems more interpretable, reliable, and effective.
• Optimize production kernels end to end by formulating optimization problems, running search loops, analyzing bottlenecks, debugging generated implementations, and landing improvements into production.
• Design abstractions, interfaces, and automation systems that accelerate kernel optimization, correctness validation, and hardware-software co-design.
• Improve AI-assisted optimization systems for specialized tasks through better datasets, evaluations, benchmarking, and research infrastructure.
• Partner across research and engineering teams to turn new ideas into practical systems spanning production needs and long-term infrastructure strategy.

Requirements

• Strong systems or tooling engineering experience, with a background in low-level software, performance optimization, or infrastructure.
• Experience with developer tooling, debugging infrastructure, profiling, observability, or workflow design for technical users.
• Depth in kernel development, accelerator architecture, compiler systems, or related performance-critical domains.
• Familiarity with AI-assisted systems, agentic workflows, post-training, or reinforcement learning for engineering or research applications.
• Strong experimental judgment, comfort with ambiguity, and the ability to move fluidly between research exploration and production execution.
• Interest in compilers, DSLs, program synthesis, or AI for systems.

Preferred

• Strong systems and tooling engineer with real depth in kernels and accelerators.
• Comfortable working across software and hardware boundaries, can reason deeply about performance, abstractions, and system design.
• Hands-on experience optimizing code for GPUs, high-performance CPUs, or custom accelerators.
• View AI not as the end product, but as a force multiplier for engineering productivity and system optimization.