Software Engineer (GPU Kernels)

• 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
• Design and implement high-performance GPU kernels for key ML operations, including matrix multiplications, attention mechanisms, and mixture-of-experts routing
• Write and optimize code using CUDA, PTX assembly, and architecture-specific techniques
• Apply advanced performance optimization methods such as memory coalescing, warp-level programming, tensor core acceleration, and compute/memory overlap
• Implement cutting-edge features like quantization (FP8/FP4), sparsity, and compute/communication overlap
• Identify and resolve performance bottlenecks using tools like Nsight Systems, Nsight Compute, and Torch Profiler
• Collaborate with research teams to productionize theoretical advancements
• Contribute to internal and open-source GPU libraries
• Present technical contributions at industry conferences (e.g., NVIDIA GTC, AWS re:Invent)

Benefits

• Remote-first work environment. The Baseten team is welcome to work from wherever they want; fully remote, in our San Francisco office, or a mix of both. Today, our team (including our founding team) is spread across the United States, Canada, and Armenia. We provide a $1,000 stipend for you to make your home-office comfortable and productive
• Regular in-person team summits. We get together as a team three times a year to plan, workshop, and most importantly, get to know each other better
• Unlimited PTO. We ask that everyone take at least 4 weeks of vacation. And we have a company-wide break between Christmas and New Year’s Day
• Full healthcare coverage. Medical, dental and vision insurance for you and your family
• Paid parental leave. 16-weeks fully paid parental leave (adoptive and non-birth parents included) and flexibility with schedules while returning to work
• Company-sponsored 401(k) for you to contribute to
• Learning and development budget. We encourage you to take classes, attend conferences, and invest in your craft and we’ll cover expenses to make it happen

• Thread/block/grid organization
• Proficient in C++ and GPU performance profiling tools
• Modern GPU features (e.g., tensor cores, async operations)
• Synchronization techniques and race condition mitigation
• CUDA C++ API
• Numerical precision and quantization strategies
• Memory access patterns and bandwidth optimization
• Memory hierarchy (global, shared, registers, L1/L2 cache)
• Strong understanding of GPU architecture and programming paradigms:
• Experience with Transformer models and attention optimization (e.g., Flash Attention)
• Background in GEMM tuning and distributed/multi-GPU compute
• Familiarity with GPU kernel libraries: Cutlass, Triton, Thrust, CUB
• Contributions to open-source GPU projects
• Research publications or conference presentations on GPU performance