Responsibilities

• Design, deploy, and maintain large distributed ML training and inference clusters

• Develop efficient, scalable end-to-end pipelines to manage petabyte-scale datasets and model training throughout the entire ML lifecycle
• Research and test various training approaches including parallelization techniques and numerical precision trade-offs across different model scales
• Analyze, profile and debug low-level GPU operations to optimize performance
• Stay up-to-date on research to bring new ideas to work
• Work across the full ML stack (data, model, eval, and infrastructure)
• Implement novel model architectures and training algorithms
• Build data pipelines and training infrastructure for massive, petabyte-scale, multimodal datasets
• Rapidly iterate on experiments and ablations
• Stay up-to-date on research to bring new ideas to work

What we’re looking for

We value a relentless approach to problem-solving, rapid execution, and the ability to quickly learn in unfamiliar domains.

• Strong grasp of machine learning fundamentals, and depth in at least one core domain (e.g. Computer Vision, Sensor Fusion, Language Models, Physics- informed NNs)
• Experienced at training models and understanding experiment results through careful analysis and ablation studies.
• Experienced at writing and optimizing massive petabyte-scale data pipelines.
• Familiarity with distributed training.
• [bonus] Familiarity with meteorology, computational fluid dynamics, and/or

numerical simulations.

• Strong grasp of state-of-the-art techniques for optimizing training and inference workloads
• Demonstrated proficiency with distributed training frameworks (e.g. FSDP, DeepSpeed) to train large foundation models
• Knowledge of cloud platforms (GCP, AWS, or Azure) and their ML/AI service offerings
• Familiarity with containerization and orchestration frameworks (e.g., Kubernetes, Docker)
• Background working on distributed task management systems and scalable model serving & deployment architectures
• Understanding of monitoring, logging, observability, and version control best practices for ML systems