Global Compute Architect & Platform Strategy Leader

lumalabs-ai; San Francisco (CA); On-site; USD 250,000 - 450,000; Full time

Job summary

LumaLabs AI in San Francisco is seeking a senior Compute leader to own the company’s global compute footprint, bridging capacity strategy, capital allocation, and systems architecture. You will design scaling roadmaps from silicon up, ensuring uninterrupted runway for research and robotics teams to ship frontier world models.

As a member of the executive team, you will turn capital into capability, overseeing the platform, vendor partnerships, and the multi-year roadmap to deliver scalable

Qualifications

• 10+ years of engineering leadership in large-scale distributed systems.
• Deep fluency in high-performance cluster topology and distributed training economics.
• Direct operational oversight of 10k+ accelerator environments.

Responsibilities

• Architect multi-year compute strategy and capacity roadmap.
• Direct the platform org including infra, datacenters, and operations teams.
• Maximize MFU and optimize fleet utilization for flagship training runs.
• Negotiate and manage large capital deployments with Finance.
• Unify global capacity to support world-model training and real-time inference.
• Interface with NVIDIA, AMD, hyperscalers and frontier silicon vendors.

Skills

Compute strategy leadership

Capacity planning

Vendor partnerships

High-performance distributed systems

Budget & capital allocation

Tools

InfiniBand/RoCE

GPU topology

On-prem vs cloud strategy

Job description

LumaLabs AI in San Francisco is seeking a senior Compute leader to own the company’s global compute footprint, bridging capacity strategy, capital allocation, and systems architecture. You will design scaling roadmaps from silicon up, ensuring uninterrupted runway for research and robotics teams to ship frontier world models.

As a member of the executive team, you will turn capital into capability, overseeing the platform, vendor partnerships, and the multi-year roadmap to deliver scalable