CUDA Kernels Engineer

About the company

Unlock the future of biological research with a pioneering AI-driven approach. This revolutionary technology integrates multiscale foundation models to predict, simulate, and program biology at every level. From DNA to protein structures, it offers unparalleled precision and insights, transforming drug design, bioengineering, and personalized medicine. Join the movement in pioneering the next era of medical research and innovation, where the power of AI meets the complexity of life itself.

Key Responsibilities

• Design and implement high-performance GPU/CPU kernels, balancing trade-offs to optimize overall hardware efficiency.
• Leverage deep understanding of hardware architecture and performance traits to drive aggressive system-level optimizations.
• Collaborate with platform teams to integrate kernels, maintain training system uptime, and identify areas for performance improvement.
• Create low-precision algorithms that maintain high throughput with minimal impact on machine learning accuracy.
• Partner with ML engineers to co-develop model architectures optimized for efficient training and inference.
• Engage with hardware vendors to provide guidance on hardware/software co-design strategies.

Job Requirement

• Proficient in C/C++ and Python, with strong coding and debugging skills.
• Deep understanding of GPU, CPU, or other AI accelerator architectures.
• Experienced in writing and optimizing compute kernels using CUDA or similar frameworks.
• Familiar with large language models (LLMs) and other foundation model architectures, as well as their training infrastructure.
• Skilled in improving ML accuracy using low-precision formats.
• Possess 1+ years of relevant industry experience.
• Motivated by incremental performance gains and optimization challenges.
• Suitable for a Senior Engineer role with a Ph.D. and 1-3 years of experience, or an Engineer role with an M.S. and 3-5 years of experience.