Software Engineer - GPU Kernels

Baseten; United States; Remote; USD 140,000 - 210,000; Full time

Job summary

Baseten in the United States is seeking a GPU Kernel Engineer to advance AI acceleration. You will design high-performance kernels for ML workloads, targeting low-latency inference across production systems.

You will work with research teams to translate breakthroughs into optimized code, using CUDA C++ and advanced profiling tools. This role offers growth and impact in a fast-paced environment serving millions of users.

Qualifications

• Strong understanding of GPU architectures and parallel programming.
• Proficient in CUDA C++ and performance profiling tools.
• Experience with memory hierarchy and bandwidth optimization.
• Familiar with tensor cores and quantization strategies.

Responsibilities

• Design and implement high-performance GPU kernels for ML operations.
• Write and optimize CUDA/PTX code for matrix multiplies and attention.
• Apply memory coalescing, warp-level programming, and overlap optimization.
• Work with researchers to productionize advancements.
• Present technical contributions at industry conferences.

Skills

GPU architecture

CUDA C++

Performance profiling

Memory optimization

Parallel programming

Tensor cores

Tools

Nsight Systems

Nsight Compute

Torch Profiler

CUDA Toolkit

Job description

ABOUT BASETEN

Baseten powers mission-critical inference for the world's most dynamic AI companies, like Cursor, Notion, OpenEvidence, Abridge, Clay, Gamma and Writer. By uniting applied AI research, flexible infrastructure, and seamless developer tooling, we enable companies operating at the frontier of AI to bring cutting-edge models into production. We're growing quickly and recently raised our $1.5B Series F, led by Altimeter Capital, Conviction Partners, and Spark Capital. Join us and help build the platform engineers turn to to ship AI products.

THE ROLE

We’re seeking a GPU Kernel Engineer to join our team at the cutting edge of AI acceleration, where your code directly impacts the performance of state-of-the-art machine learning models. As a GPU Kernel Engineer, you'll craft the foundation that powers modern AI workloads, optimizing every microsecond of computation to enable breakthrough applications. You'll work in a fast-paced, intellectually stimulating environment where technical excellence is paramount and your contributions directly influence production systems serving millions of users across numerous products. This role offers exceptional growth potential for engineers passionate about low-level optimization and high-impact systems work.

EXAMPLE INITIATIVES

You’ll get to work on these types of projects as part of our Model Performance team:Baseten Embeddings Inference: The fastest embeddings solution availableThe Baseten Inference StackDriving model performance optimization

RESPONSIBILITIES

Core Engineering ResponsibilitiesDesign and implement high-performance GPU kernels for key ML operations, including matrix multiplications, attention mechanisms, and mixture-of-experts routingWrite and optimize code using CUDA, PTX assembly, and architecture-specific techniquesApply advanced performance optimization methods such as memory coalescing, warp-level programming, tensor core acceleration, and compute/memory overlapPerformance * InnovationImplement cutting-edge features like quantization (FP8/FP4), sparsity, and compute/communication overlapIdentify and resolve performance bottlenecks using tools like Nsight Systems, Nsight Compute, and Torch ProfilerCollaborate with research teams to productionize theoretical advancementsImpact * CollaborationContribute to internal and open-source GPU librariesPresent technical contributions at industry conferences (e.g., NVIDIA GTC, AWS re:Invent)

REQUIREMENTS

Strong understanding of GPU architecture and programming paradigms:Memory hierarchy (global, shared, registers, L1/L2 cache)Thread/block/grid organizationSynchronization techniques and race condition mitigationProficient in C++ and GPU performance profiling toolsKnowledge of:CUDA C++ APIMemory access patterns and bandwidth optimizationNumerical precision and quantization strategiesModern GPU features (e.g., tensor cores, async operations)

NICE TO HAVE

Experience with Transformer models and attention optimization (e.g., Flash Attention)Familiarity with GPU k