Tech Lead, Gemini Inference Performance, DeepMind

DeepMind is seeking a Tech Lead for its Gemini Inference Performance team in London. The role involves setting the technical direction for performance engineers, managing team members, and prioritizing optimization tasks. Daily responsibilities include conducting roofline analysis, hardware profiling, and systems analysis to address performance issues across machine learning frameworks and custom kernels. The position requires a strong understanding of Transformer and Mixture-of-Experts models to enhance execution efficiency and memory usage. Collaboration with research teams is essential to assess the impact of architectural decisions on latency and costs. Candidates should possess a Bachelor’s degree or equivalent experience, along with at least eight years in software development and five years in a leadership role. The role is suited for individuals with a passion for AI and a drive to solve complex challenges through teamwork. The job's specific working arrangement is not mentioned.