Staff GPU Compute Infrastructure Engineer

Causal; San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Causal is building a Large Physics foundation Model and seeks an infrastructure engineer to design, deploy, and operate large distributed GPU clusters. You will extend schedulers, build self-serve interfaces, and own storage and lineage for checkpoints and logs.

You will collaborate with researchers to optimize performance and placement, ensuring reliable, scalable compute for rapid research iteration.

Qualifications

• Experience operating large-scale GPU clusters and container orchestration systems.
• Strong systems background: Linux, networking, storage, infra-as-code.
• Knowledge of cloud platforms (GCP, AWS, Azure) and ML/AI services.
• Understanding of monitoring, logging, observability, and version control for ML.
• Familiarity with CUDA/NCCL and distributed workload profiling.
• Owns deliverables end-to-end, from requirements through autonomous execution.

Responsibilities

• Design, deploy, and operate large distributed GPU clusters end to end.
• Extend scheduling and orchestration for multi-tenancy across workloads.
• Build software that abstracts cluster management for researchers and engineers.
• Own cluster storage and artifact paths with retention and lineage.
• Monitor and improve reliability and error recovery; build observability.
• Partner with researchers to unblock large-scale runs and advise on placement.

Skills

GPU clusters operations

Kubernetes

Slurm

Docker

Linux

Networking

Infrastructure as code

Cloud platforms (GCP, AWS, Azure)

Monitoring & observability

Version control for ML

CUDA/NCCL

Performance profiling

End-to-end ownership

Tools

Kubernetes

Slurm

Docker

Job description

Causal is building a Large Physics foundation Model and seeks an infrastructure engineer to design, deploy, and operate large distributed GPU clusters. You will extend schedulers, build self-serve interfaces, and own storage and lineage for checkpoints and logs.

You will collaborate with researchers to optimize performance and placement, ensuring reliable, scalable compute for rapid research iteration.