Senior Site Reliability Engineer

Team: Infra Reliability · SF Bay Area / Remote (US) You'll own the GPU infrastructure Luma's research and product run on — thousands of NVIDIA and AMD GPUs across on-prem and multi-cloud (AWS and OCI). As a Senior SRE, you keep training and inference clusters reliable and fast, and you help redesign them for the next level of scale. This is a hands-on, close-to-the-metal role for a first-principles Linux engineer. You'll be the final escalation for the hardest GPU, networking, and kernel-level failures, sometimes debugging directly with NVIDIA. It fits someone who thrives on low-level problems in a fast, less-structured environment. If you want a narrow, well-bounded ops role, this isn't it. What You'll Own - Take end-to-end ownership of production GPU clusters for training and inference across AWS and OCI, keeping them highly available and performant. - Join critical re-architecture sessions to redesign systems for higher efficiency and scale. - Tune Linux performance deeply, at the OS and kernel level. - Build automation in Python, Go, or Bash to manage, monitor, and self-heal infrastructure without heavy toil. - Serve as the final escalation for the hardest GPU, networking (InfiniBand/RDMA), and system failures, working with vendors like NVIDIA. - Help achieve and maintain security certifications (SOC 2 Type 1 & 2, ISO) with strong infrastructure security practices. First 90 Days One way the first 90 could unfold. - Days 1–30 — Immerse & Diagnose: Learn the current clusters across on-prem, AWS, and OCI, and where reliability and performance hurt most. - Days 30–60 — Ship & Validate: Take ownership of a production cluster and ship automation or tuning that measurably improves availability or performance. - Days 60–90 — Scale & Systemize: Contribute to the next-gen re-architecture and harden security and compliance practices. What You Bring - 5+ years as an SRE, production, or infrastructure engineer in a fast-paced, large-scale environment. - Deep, hands-on Linux expertise, containerized systems, and low-level performance debugging. - Working experience with Terraform, Airflow, and Ray. - Strong experience with AWS or OCI. - Practical experience with high-performance networking (InfiniBand, RDMA, or RoCE). - Working knowledge of security best practices and compliance frameworks like SOC 2 and ISO. - Comfort in a less-structured, fast-paced environment. Nice to Have - Deep expertise with GPU tooling for NVIDIA and AMD (DCGM, ROCm). - Experience managing large-scale GPU clusters for AI/ML training or inference. - Familiarity with Kubernetes or orchestration frameworks like Ray. - Deep expertise in data pipelines and infrastructure. About Luma: Luma's mission is to build unified general intelligence that can generate, understand, and operate in the physical world. We believe multimodality is critical for intelligence — the next step beyond language models comes from vision. Luma is an equal opportunity employer.