Kernel Engineer

Recursive Superintelligence • San Francisco, CA

2d

About The Position

The Kernel Engineer improves the speed, efficiency, and reliability of machine learning training and inference at the hardware–software boundary. The role is suited to engineers with deep implementation-level expertise in at least one of GPU or accelerator kernels, ML compilers and accelerator backends, or framework, runtime, and inference performance; experience across all three areas is not expected.

The engineer develops and optimizes performance-critical components, identifies bottlenecks using profiling and measurement, and improves ML training and inference stacks while maintaining correctness and numerical reliability. Depending on expertise, the work may involve kernels, compiler and code-generation pipelines, framework backends, execution runtimes, or serving systems. The role works with technologies such as PyTorch, JAX, vLLM, and SGLang where relevant, and involves collaboration with researchers, infrastructure engineers, and hardware-aware systems experts.

Success means uncovering important performance limitations, delivering technically sound and validated improvements with clear before-and-after evidence, and building trust through rigorous measurement, independent technical judgment, and effective collaboration.

Requirements

• At least two years of post-graduation industry experience in GPU or accelerator kernels, ML compilers or backends, framework or runtime systems, inference performance, or comparably low-level infrastructure.
• Deep implementation-level expertise in at least one relevant stack, such as CUDA/HIP, Triton, Pallas, NKI/Neuron, XLA, MLIR, TVM, PyTorch/JAX backends, TensorRT, vLLM, or SGLang.
• Ownership of substantive kernel, compiler, backend, or runtime improvements, beyond API usage; ability to explain design decisions, technical contributions, correctness checks, and measured impact or credible experimental evidence.
• Strong understanding of hardware execution and performance fundamentals, including memory hierarchy, parallelism, latency, throughput, and performance profiling.
• Ability to reason about numerical correctness, performance regressions, and tradeoffs involved in optimizing ML workloads.
• Evidence of delivering complex technical work collaboratively, from investigation through implementation and validation.

Nice To Haves

• Meaningful contributions to ML frameworks or inference and serving systems, such as PyTorch, JAX, vLLM, or SGLang, with clear ownership of a performance or systems outcome.
• Experience with GPU software, TPU, Trainium, or other custom accelerator stacks; kernel libraries, compiler infrastructure, or chip-adjacent software.
• Low-level systems engineering in high-performance infrastructure, even if previous work was not directly focused on large language models.
• Research that translates into engineering results, with clearly distinguished individual contributions and practical impact.

Responsibilities

• Develop and optimize performance-critical components across GPU and accelerator kernels, ML compiler and code-generation pipelines, or framework and serving runtimes.
• Identify bottlenecks across compute, memory movement, scheduling, and execution, using profiling and measurement to guide optimization.
• Improve the performance, efficiency, or capabilities of machine learning training and inference stacks, including kernels, runtime components, and compiler pathways.
• Work with frameworks and serving systems such as PyTorch, JAX, vLLM, and SGLang, where relevant, to translate workload needs into low-level systems improvements.
• Maintain correctness and numerical reliability while making performance tradeoffs; build clear benchmarks and validation approaches.
• Collaborate with researchers, infrastructure engineers, and hardware-aware systems experts to scope problems, ship improvements, and complete projects.
• Develop and tune compute primitives, memory layouts, fusion, parallelism, and numerical precision for high-performance execution, where relevant.
• Improve compiler lowering, code generation, scheduling, graph transformations, and hardware-specific execution, where relevant.
• Optimize GPU backends, execution runtimes, serving systems, model operators, and end-to-end training or inference bottlenecks, where relevant.

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level
• Education Level: No Education Listed