Member of Technical Staff (GPU Engineer)

As a GPU Engineer at Reka, you'll be responsible for designing and implementing enhancements to our training infrastructure. This includes optimizing model training performance and implementing robust monitoring systems for performance tracking. You will also work on writing and debugging low-level GPU code to ensure efficiency and reliability within our system.

To be a suitable candidate for the GPU Engineer role at Reka, you should possess strong engineering skills in Python, along with experience in PyTorch or other acceleration libraries. Additionally, familiarity with low-level GPU programming using CUDA or Triton is crucial, alongside experience in managing large GPU compute clusters using tools like Slurm or Kubernetes.

At Reka, we utilize a variety of tools and technologies tailored for GPU engineering including Python, PyTorch, CUDA, and Triton for low-level programming. We also leverage large-scale compute clusters and container orchestration systems like Slurm and Kubernetes to efficiently scale GPU jobs across our infrastructure.

The work culture at Reka is inclusive and open, with a strong emphasis on collaboration among a diverse group of engineers, researchers, and operators. We promote a mission-focused environment that encourages innovative ideas and fosters creativity, making it an ideal place for a GPU Engineer to thrive and contribute meaningfully to the field of artificial intelligence.

Yes, Reka embraces a remote-first approach, allowing our GPU Engineers and other team members to work from anywhere in the world. This setup not only helps us bring together top-tier talent but also supports a healthy work-life balance.

In your response, focus on specific projects where you've leveraged Python for GPU programming. Discuss how you've utilized libraries like PyTorch or TensorFlow, and mention any experience with performance optimization for model training.

Detail your experience with tools such as Slurm or Kubernetes. Talk about specific strategies you've implemented to optimize job scheduling and resource allocation in a large-scale environment.

Share your methods for diagnosing and fixing bugs within low-level GPU code. Include examples of tools and techniques you've used, like CUDA debugging tools or hardware profiling techniques.

Discuss your experience with monitoring systems and performance tracking. Highlight any tools or practices you have implemented to detect anomalies and maintain optimal efficiency during training cycles.

Provide details about your experience with Ceph or similar filesystems, focusing on how you've integrated them into GPU computing environments and their impact on system performance and data management.

Explain your commitment to continuous learning. Mention any relevant conferences, workshops, or online courses you pursue, as well as valuable resources like research papers and industry publications you follow.

Illustrate your approach to cross-team collaboration. Discuss specific tools you've used for communication and project management, and provide examples where teamwork helped drive a successful outcome.

Explain the situation, the options you considered, and the outcome of your decision. Highlight how it affected the project's success and what you learned from the experience in the context of GPU engineering.

Discuss a scenario where you had to optimize for performance while considering trade-offs. Highlight your decision-making process and the results of your optimizations.

Share your aspirations in the field of GPU engineering, emphasizing your desire to contribute to cutting-edge AI technologies and improve model training performance. Express your commitment to innovation and your interest in leadership roles within the industry.