Carbon-Aware Data-Centre Scheduling Director — Cloud Infrastructure

The mandate

The operator reports annual renewable matching while its fastest-growing AI and batch workloads increasingly run during locally constrained grid periods. Engineering can shift some jobs across time and region, but current carbon signals disagree, customer latency boundaries are poorly classified and energy contracts obscure marginal consequences. Leadership needs a control design that makes defensible operational choices, not a retrospective emissions allocation.

The deliverable is a Carbon-Aware Scheduling Control Blueprint covering eligible workload classes, grid and emissions signals, forecast uncertainty, orchestration rules, customer constraints, accounting treatment and operator overrides. It will include a production-shadow prototype, counterfactual results, service-risk assessment and investment case. Claims must distinguish locational and temporal operating effect from certificates or annual contractual matching.

Milestone one in week three establishes signal provenance, workload taxonomy and decision baseline. Week seven concludes milestone two with scheduling policies and a replay simulator. By week twelve, milestone three supplies shadow-operation evidence from two Nordic sites. The final blueprint, assurance record, customer-language recommendations and scale decision reach the joint sponsors in week seventeen.

Acceptance requires Sustainability Assurance to reproduce the avoided-emissions calculation from source intervals, Infrastructure to show that shadow schedules respect tested latency and recovery constraints, and Finance to reconcile energy and capacity effects. At least three unseen grid scenarios must produce explainable decisions within approved bounds. The sponsors must approve both the eligible-workload perimeter and conditions under which carbon optimisation is overridden.

The client will provide interval energy use, power-market and grid data, contractual instruments, workload metadata, scheduler logs, capacity forecasts, service terms and incident histories. Platform engineers will operate the shadow environment; Legal will review customer commitments; and an energy specialist will document signal limitations. The infrastructure chief will resolve access and policy disputes within two working days.

Why this is external work

Sustainability teams understand reporting and procurement, while schedulers optimise service and capacity; neither owns a neutral causal test of marginal emissions. Several suppliers also promote proprietary signals linked to their broader services. External work provides cross-domain arbitration and an evidence standard robust enough for engineering, finance and assurance to accept simultaneously.

What you will own

• Classify workloads by deadline, latency, locality, checkpointability, data sovereignty, recovery need and customer permission before testing flexibility.
• Reconcile grid carbon, marginal generation, market price, curtailment and contractual signals, documenting provenance, interval mismatch and forecast error.
• Design scheduling policies that expose the priority order among safety, service, capacity, cost and emissions rather than hiding trade-offs in weights.
• Build a replay and shadow method comparing proposed decisions with actual operation across congestion, renewable surplus and forecast-failure scenarios.
• Quantify emissions, energy-cost, capacity and service effects with separate causal claims for temporal shift, regional move and reduced computation.
• Define operator overrides, decision logging, customer notification and assurance evidence for cases where the lowest-carbon action is not permissible.
• Transfer the simulator, signal qualification, policy-change process and monthly effectiveness review to named infrastructure and sustainability owners.

Candidate qualifications

• Designed workload orchestration or energy controls for hyperscale, colocation, HPC or large industrial computing estates exposed to variable power systems.
• Combined interval grid or marginal-emissions evidence with workload constraints rather than relying solely on annual renewable-energy matching.
• Validated scheduling policy through replay or shadow operation and can quantify both avoided emissions and service or capacity consequences.
• Worked across energy procurement, distributed systems, finance and sustainability assurance where each function used different boundaries and measures.
• Challenged a carbon claim whose certificate accounting was valid but did not demonstrate a changed physical operating decision.
• Delivered control logic and evidence that internal engineers could operate after engagement closure without dependency on proprietary consultancy tools.

Non-negotiables

• The named director must lead technical and assurance workshops in Copenhagen and spend scheduled validation periods at the Jutland sites.
• No commission, data resale arrangement or investment interest may exist with energy-signal, orchestration, certificate or data-centre vendors considered.
• Customer service, safety and data-sovereignty constraints cannot be silently relaxed to improve the modelled carbon result.
• All public-claim language remains subject to client legal and assurance approval and sits outside technical acceptance of the scheduling blueprint.

1. 49 words maximum. Describe one workload shift where interval grid evidence changed the decision and quantify its service consequence.
2. 49 words maximum. How would you distinguish avoided physical emissions from a valid but unchanged renewable-contract allocation?
3. 49 words maximum. Which client data is indispensable for testing carbon-aware scheduling without breaching latency or sovereignty commitments?