GPU Kernel Engineer
Location: Hybrid (preferably SF or Vienna), San Francisco, CA
Salary: Competitive Salary, Competitive Equity
Job Type: Full-time
Responsibilities:
- Implement high-throughput low-latency GPU kernels in CUDA or Triton
- Efficient mixed-precision training and inference
- Optimize kernel fusions
- Profile and benchmark competing implementations
- Maximize multi-GPU communication through efficient overlapping of communication and computation
- Ensure the compatibility of GPU driver software with the latest operating systems
- Develop and maintain GPU driver software
- Develop and implement high-performance algorithms and data structures for GPU processing
- Collaborate with other engineers to debug and resolve software issues
- Continuously learn new GPU technologies and architectures
- Write technical documentation and provide feedback on product design.
Requirements:
- At least 8 years of relevant software development experience
- 5+ years of experience in GPU kernel development and optimization
- MS or PhD in Computer Science, Electrical Engineering, or related field
- Strong proficiency in C and C++ programming languages
- Strong programming skills in CUDA and C++
- Experience with GPU hardware and programming models (e.g., CUDA, OpenCL, Metal, Vulkan)
- Experience with GPU profiling and performance analysis
- Experience with deep learning frameworks such as TensorFlow or PyTorch
- Familiarity with GPU architectures, including memory hierarchy and parallel computing
- Knowledge of software development best practices, including debugging, testing, and version control
- Excellent problem-solving and analytical skills
- Strong communication and collaboration skills.
Benefits and Perks:
- Benchmark-based compensation in the 75th or 90th percentile, including base salary, equity, and benefits
- Flexible working hours
- In-person (SF or Vienna) or remote work options
- A small, fast-paced, highly focused team.
The Ideal Candidate:
We are looking for an ambitious and high-energy individual who is aligned and mission-driven. The ideal GPU Kernel Engineer will work on ensuring high throughput and low latency during training and inference with giant neural networks. While experience is important, we value a budding superstar over an 8-year experienced engineer with a PhD whose experience looks perfect on paper but may not be the right fit for our team. We want a candidate who is joining a mission-driven team that’s well-equipped with the best talent to build aligned and more complete AI to accelerate humanity’s progress on the world’s most important problems.
To apply, the interested candidate may submit his/her updated English resume to the email address: careers [ at ] somosotech.com or by clicking the "apply now" button below.
-------