Software Engineer, Kernels & Performance

About the role

We’re looking for a seasoned Research Engineer with expertise in GPU computing to design high-performance kernels for our training and inference workloads using CUDA/CUTLASS, C++, and Python.

• Using the capabilities of GPUs and other accelerators to the fullest extent to make our custom model architectures fast and efficient for training and inference
• Deliver and maintain high‑performance GPU and communication kernels that increase utilization and also work well in a distributed training and inference environment
• Collaborating closely with ML Engineers and Researchers to develop new architectures and algorithms that are aware of I/O & hardware constraints
• Communicating with hardware vendors to advise on the design of software or future hardware, while also making sure we are keeping up to date with the cutting edge
• Working on quantization or any form of low‑precision arithmetic to increase throughput without degradation in performance

• Strong C/C++ and Python coding skills
• Deep understanding of GPUs or other accelerators, along with the ability to effectively profile and analyze existing or new kernels
• Experience with CUDA/CUTLASS (and experience with Triton‑like compilers is a plus)
• 3+ years of relevant industry experience and experience working with hardware developers
• MS/PhD in Computer Science and Engineering with a specialization in Computer Architecture, Parallel Computing, Compilers or other system disciplines

Join our visionary team, on‑site 3‑5 days a week, at the forefront of next‑generation AGI.