Member of Technical Staff — Compute Cluster

To achieve this breakthrough, we are building a Large Physics foundation Model (LPM) because physical systems, unlike text or images, are governed by verifiable cause and effect. We believe that scaling on physics will enable an understanding of causality required to predict and control physical systems, starting with weather.

We look for infrastructure engineers who are excited to tackle unsolved problems. Everything we do — training, evaluation, serving — runs on our GPU fleet. Your mission is to design, build, and operate the supercomputing environment underneath it all, delivering performant, reliable, and cost-efficient compute to ensure research is able to iterate rapidly at scale.

Responsibilities

• Design, deploy, and operate large distributed GPU clusters end to end: provisioning, imaging, upgrades, and capacity planning
• Extend scheduling and orchestration systems (e.g. Kubernetes, Slurm) for topology-aware placement, preemption, quotas, and multi-tenancy across training and inference workloads
• Build software that abstracts cluster management and presents a unified, self-serve interface to researchers and engineers
• Own cluster storage and artifact paths for checkpoints and logs, with clear retention and lineage
• Monitor and continuously improve reliability and error recovery; build the observability to catch failures before researchers do
• Partner with researchers to unblock large-scale runs and advise on performance and placement trade-offs

What we're looking for

• Experience operating large-scale GPU clusters and container orchestration frameworks (e.g. Kubernetes, Slurm, Docker)
• Strong systems background: Linux, networking, storage, infrastructure-as-code
• Knowledge of cloud platforms (GCP, AWS, or Azure) and their ML/AI service offerings
• Understanding of monitoring, logging, observability, and version control best practices for ML systems
• Familiarity with CUDA/NCCL and performance profiling for distributed workloads
• Owns deliverables end-to-end, from requirements through autonomous execution