You'll own the reliability of Luma's 10k+ GPU fleet: the scheduling, efficiency, and resilience that research and products depend on. As a Staff AI Infrastructure Engineer, you'll be a technical authority who turns deep systems knowledge into repeatable, company-wide reliability, and a leader other strong engineers want to work with.
This is close-to-the-metal work — kernels, containers, schedulers, networking, storage, GPU behavior — under demand hard enough that yesterday's solutions break regularly. It's also a technical-leadership role: you'll set the bar and grow the team. If most of your experience has been inside highly abstracted internal platforms where others owned the underlying machinery, this likely isn't a match.
What You'll Own
Architect and operate large, heterogeneous GPU environments under extreme demand, improving utilization and performance where small gains change company outcomes.
Resolve failures spanning hardware, OS, runtimes, and orchestration, and eliminate whole classes of instability.
Define how infrastructure and workloads evolve as cluster size and concurrency grow — scheduling, placement, resource management.
Work directly with research to build the systems new model capabilities require, and scale inference without sacrificing reliability or latency.
Hire and develop exceptional systems and reliability engineers, and set the bar for depth, judgment, and production ownership.
Shape product and research architecture early through strong partnerships.
First 90 Days
One way the first 90 could unfold.
Days 1–30 — Immerse & Diagnose: Learn the fleet, its failure modes, and the biggest reliability and utilization gaps.
Days 30–60 — Ship & Validate: Eliminate a recurring class of instability or land a utilization or performance win that moves company outcomes.
Days 60–90 — Scale & Systemize: Set the reliability direction, redesign ahead of where today's abstractions will fail, and begin building the team.
What You Bring
Deep expertise in Linux and distributed systems.
Experience operating GPU or accelerator clusters in real production environments.
Strong fluency in Kubernetes and modern open-source infrastructure.
Comfort debugging across hardware, kernel, runtime, and orchestration, and understanding how systems behave under contention and at scale.
You write code and build automation, and think in bottlenecks, failure modes, and trade-offs.
Judgment engineers trust, especially when things break.
Nice to Have
You raise reliability standards company-wide and influence product and research architecture early.
You build partnerships rather than ticket queues, and attract and level up strong engineers.
Curiosity for how models use infrastructure, because improving systems expands what becomes possible.
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.