TPU Kernel Engineer
Core
Designing and optimizing kernels for TPUs to address performance issues across ML systems for research, training, and inference.
Role type
TPU Kernel Engineer
Builds
Low-latency, high-throughput sampling for large language models; custom collective communication algorithms; adapted low-precision inference models.
Domain
Machine Learning / High-Performance Computing / TPU Architecture
Deliverable
production ML models
Required skills
TPU kernel optimization, low-level optimization, large-scale ML systems, computer architecture, ML framework internals, debugging at assembly level
Preferred skills
GPU optimization, language modeling with transformers, quantitative system performance modeling
Technologies
TPU, GPUs, ML frameworks, assembly
Responsibilities
Identify and address performance issues in ML systems, design and optimize TPU kernels, provide feedback to researchers on model performance impact, implement custom collective communication algorithms, debug kernel performance at the assembly level.
Seniority
Mid-to-Senior, hands-on IC
