Senior Software Engineer, Generative AI, Google Research
Overview:
In this role you design, build, and maintain scalable performance simulation frameworks to evaluate training and serving metrics for next-generation TPU configurations. You model interactions between advanced model architectures and architectural variables to guide hardware roadmap decisions. You will automate simulation sweeps with AI-driven platforms, and develop optimization and analytical engines to identify Pareto-optimal execution strategies. You'll translate results into data-grounded projections to inform multi-year TPU platform strategies and contracts. This position sits at the intersection of frontier ML models and accelerator roadmaps, offering meaningful impact on hardware and??Compensation / Benefitsbonus targetequitybenefitscompetitive compensationopportunities for impactResponsibilitiesDesign, build, and maintain scalable performance simulation frameworks for TPU configurationsModel interactions between future model architectures and architectural variablesDevelop AI-driven platforms to automate sweep setup, parameter sensitivity analyses, and data extractionBuild analytical/optimization engines (e.g., MIP, simulated annealing, roofline) to identify Pareto-optimal strategiesFormulate data-grounded performance projections to inform TPU roadmaps and hardware contractsIntegrate LLM-driven agentic workflows to automate sweep generation, exploration, data retrieval, and bottleneck diagnosisKey requirements 5 years of software engineering experience with C++ and PythonExperience with computer architecture concepts (memory hierarchies, roofline models, bandwidth vs compute bottlenecks, network topologies) Experience in software design for simulation, performance modeling, or scientific computingleadershipadaptabilitycross-functional collaborationC++Pythonsimulation design