Tech Lead, Gemini Inference Performance, DeepMind

In this role you will set the technical roadmap for a team of performance engineers and mentor direct reports while prioritizing high-impact optimization opportunities. You will apply roofline analysis, hardware profiling, and systems analysis to remove bottlenecks across ML frameworks, compilers, custom kernels, and serving infrastructure on accelerators. You will leverage a first-principles understanding of Transformer and Mixture-of-Experts components to optimize execution efficiency and memory. You will collaborate with research teams early to anticipate inference implications and drive cost-efficient, low-latency serving solutions.

• Set and manage the technical roadmap for performance engineers
• Support and develop direct reports and high-potential engineers
• Prioritize optimization opportunities across competing bottlenecks
• Apply roofline analysis and hardware-level profiling to identify bottlenecks
• Analyze bottlenecks across ML frameworks, compilers, custom kernels, and serving infra on accelerators
• Focus on Transformer and Mixture-of-Experts components to optimize execution and memory footprint
• Collaborate with research teams to evaluate inference implications and cost/latency impact
• Contribute to development of custom kernels and serving optimizations
• Conduct deep performance profiling using hardware tracing tools to assess accelerator utilization and interconnect latency across large-scale systems

• Bachelor’s degree or equivalent practical experience
• 8 years of software development experience
• 5 years in people management/supervision
• 5 years in a technical leadership role overseeing projects

• leadership and people management
• cross-functional collaboration
• strategic planning
• roofline analysis
• hardware-level profiling
• systems analysis