Staff Software Engineer, ML Performance, GPU - Google
Software Engineer
We're looking for a Software Engineer focused on designing and building scalable technical solutions. This lead role requires 8+ years of relevant experience.
About the role
Minimum qualifications:
Bachelor’s degree or equivalent practical experience.
8 years of experience in software development.
5 years of experience with ML design and ML infrastructure (e.g., model deployment, model evaluation, data processing, debugging, fine tuning).
Experience with modern GPU architectures, memory hierarchies, and performance bottlenecks.
Experience with low-level GPU programming (CUDA, Triton, CUTLASS, etc.) and performance engineering techniques.
Experience with modern LLMs and their deployment on AI accelerators.
Preferred qualifications:
Master’s degree or PhD in Engineering, Computer Science, or a related technical field.
8 years of experience with data structures and algorithms.
3 years of experience in a technical leadership role leading project teams and setting technical direction.
3 years of experience working in a complex, matrixed organization involving cross-functional, or cross-business projects.
Experience in hardware-aware algorithm design and compiler stacks (e.g., OpenXLA), tailoring large-scale ML models and distributed systems for peak performance across accelerator hardware.
Responsibilities
Identify and maintain LLM training and serving benchmarks; use them to identify performance opportunities, drive XLA:GPU/Triton performance and guide XLA releases.
Partner with product teams (e.g., Google DeepMind) to onboard, optimize, and scale LLMs and machine learning models on GPU hardware.
Conduct architecture-level simulations, performance benchmarking, and roofline analyses using tools like TRT-LLM, vLLM, and SGLang to guide system designs.
Analyze fleet-wide performance and efficiency metrics to identify bottlenecks and engineer scalable optimizations across Google's infrastructure.
Research and implement model/data efficiency techniques, tooling, and profiling mechanisms to improve workload performance and training efficiency.
Skills
machine learningnatural language processingllmcuda