Senior Software Engineer, ML Performance, GenAI, DeepMind
About The Position
At Google DeepMind, the mission is to build the world's first general-purpose learning agent. This role focuses on measuring and improving the performance of production and research machine-learning workloads, particularly LLM inference, including serving latency, throughput, and cost per token, as well as agentic workflows on current and next-generation TPU platforms. The work spans model architecture and sharding strategies through Pallas/Mosaic kernels and XLA, and involves close collaboration with Gemini modeling, Deployment, Compiler, and TPU hardware teams to shape next-generation accelerator capabilities. Google DeepMind is a pioneering AI lab with interdisciplinary teams advancing AI to address complex global challenges and support product innovation. The organization emphasizes widespread public benefit, scientific discovery, and safety and ethics. The role offers varied learning opportunities and career pathways in a collaborative, mission-driven environment.
Requirements
- Bachelor's degree in Computer Science, Software Engineering, Mathematics, or a related STEM field, or equivalent practical experience.
- At least 5 years of experience in software development, systems programming, or applied ML engineering, or equivalent education credit such as internships, student research, or pre-graduate work experience.
- At least 3 years of experience with performance optimization, systems data analysis, visualization tools, or debugging.
- At least 3 years of experience with Generative AI, Large Language Models (LLMs), Machine Learning, and related frameworks.
- At least 1 year of experience designing and launching production software products or foundation AI systems.
Nice To Haves
- Master's or PhD in Computer Science, Artificial Intelligence, Robotics, or a related technical field.
- At least 3 years of experience in a technical leadership role setting architectural direction and guiding project teams.
- Experience developing, evaluating, or serving large language models, agentic workflows, multimodal pipelines, or AI runtime APIs such as PyTorch, JAX, Triton, vLLM, or Gemini API.
- Expertise or strong interest in performance debugging and optimization of complex workloads, design of performance tools, compiler design and code optimization, high-performance software development techniques, concurrent programming, or multi-core computer architectures.
Responsibilities
- Reduce Gemini inference cost per token by optimizing throughput, accelerator utilization, batching, scheduling, and KV cache efficiency, including building agentic workflows that help AI agents discover, implement, and validate performance improvements at scale.
- Bring Gemini to new TPU generations by characterizing hardware at launch, tuning kernels, sharding, and serving configurations for specific memory hierarchies and interconnects, and delivering production-ready performance.
- Share hardware and performance insights to influence future TPU hardware and compiler roadmaps.
- Implement model-compression techniques, such as quantization and sparsity, to reduce the cost per token for production Gemini models.
- Evaluate new architectures, optimizers, and training methods to achieve target quality with reduced compute requirements.
Benefits
- 15% bonus target
- Equity
- Benefits
Stand Out From the Crowd
Upload your resume and get instant feedback on how well it matches this job.
What This Job Offers
Job Type
Full-time
Career Level
Senior
