Crusoe is the world's first vertically integrated, sustainable AI cloud. We build and operate GPU infrastructure powered by clean energy, from data center design through IaaS products to managed inference at scale, enabling AI-native companies to run demanding workloads without compromising on sustainability or reliability. Crusoe Cloud is 1,400 people and growing, and the TPM frameworks are still being built -- which means there is a genuine opportunity to shape how the function operates rather than inherit how it already works
The Managed Inference platform is where customers run production LLM workloads without managing low-level infrastructure, and it is one of Crusoe's fastest-growing product areas. The Staff TPM for Managed Intelligence connects model engineering, IaaS, product, and data center operations to deliver a reliable, scalable inference platform. You will own end-to-end program delivery across multi-quarter roadmaps, model onboarding, inference optimization, and production readiness for new model versions. Deep familiarity with the model layer -- including how LLMs are served, optimized, and evaluated in production -- is essential to being effective in this role
End-to-end program delivery: Own multi-quarter release planning, dependency governance, and executive communication across the Managed Inference platform
Complex, high-risk program management: Drive model version rollouts, inference optimization campaigns, SLA readiness for new GPU hardware, and multi-tenant capacity planning from kickoff through delivery
Cross-functional alignment: Coordinate across Model Engineering, IaaS, Cloud Foundations, Data Center Operations, and external model providers to keep programs on track and unblocked
Proactive risk identification: Surface risks across model serving, reliability, capacity constraints, and vendor timelines before they become program-level problems
Execution frameworks and dashboards: Build lightweight, scalable TPM frameworks suited to Crusoe's pace; maintain real-time execution dashboards and deliver crisp, data-driven executive updates
Phase 0 planning for model onboarding: Own pre-launch planning for model onboarding on new GPU generations, including firmware and driver readiness, CUDA and ROCm stack validation, and commissioning criteria for inference workloads
Stakeholder leadership: Drive alignment and push back effectively across engineering, product, and operations leadership -- including highly technical stakeholders who have not previously worked with a TPM