Power, not silicon, is the binding constraint on AI infrastructure. Crusoe is vertically integrated from electrons to tokens; generation, batteries, controls, and the cloud orchestration layer which means we can credibly commit flexible load to utilities and deliver on it. No other AI cloud can
Flexible Compute is the system that makes that real: reducing data-center power draw on demand quickly, reliably, verifiably — while protecting the workloads that matter most. It gets us megawatts sooner, earns more per megawatt, and rides through grid-stress events without breaking SLAs
You'll operate at the seam between one of the industry's fastest-growing energy footprints and the software that controls it. This is a utility-facing, safety-critical control system, and a ground-up build. You’ll define the architecture, ship the first production system, and build the team from scratch
Own the curtailment orchestration and decision layer: grid-signal ingestion, staged shedding, per-SKU power capping without ever silently dropping a paid workload
Land a utility-validated pilot: fast ramp to setpoint, tight accuracy, high-fidelity telemetry, and the test harness that proves it before we commit to a utility
Deliver dynamic power management for oversubscription — more GPUs per megawatt with no customer-visible impact
Build fast, workload-aware GPU power estimation, validated against fleet telemetry : the engine behind shed forecasting, oversubscription admission, and power planning for new silicon
Design for safety: authenticated signal ingress, bounded blast radius, fail-safe defaults, manual backstops
Own graceful ride-through of power-loss and grid-stress events, integrated with on-site battery and generation backstops
Partner deeply with Data Center Engineering (mechanical, electrical, controls) and Energy teams on interconnection commitments, curtailment program design, and BESS/generation integration
Integrate with our cloud control plane across Kubernetes and Slurm fleets : one control plane, not two
Make build-vs-leverage calls across vendor power-management stacks and grid integration layers
Hire and lead the team from the ground up, keeping headcount sublinear to fleet growth through automation