Data Center Operations Manager for AI and GPU Cloud Providers

What does this hire really come down to?

At a neocloud the operations manager sits where facility engineering meets high performance computing. Customers sign for dense racks and expect their jobs to run for weeks. A fault that a traditional colocation customer might shrug off can cost a training run. The manager must know the plant and the cluster well enough to talk to both sides without translation.

The tradeoff is process against pace. Managers from mature enterprise sites bring change control, audit habits and steady shifts, but may balk at a schedule that changes weekly. Managers from startups or high performance computing bring speed, and sometimes carry loose habits around change records. Customer contracts and audit demands decide which risk you can accept.

Retrofits add a third problem. Many providers fit liquid cooling into halls built for air. The manager has to schedule that work around running customers. Ask who has done a retrofit while neighbors stayed live, and what they would change. Funding matters here more than elsewhere. Stable managers read the company's runway before they read the job. Ask how the manager will be supported on audits. If your compliance team is one person, the manager will carry more than the job description says.

What do we check before you see anyone?

Ask how they handled a customer service level breach, including what they told the customer and what changed in the plant afterward. Ask which audit frameworks they supported and what evidence they personally gathered, since support in name only is common.

• Ask how they handled a customer service level breach, including what they told the customer and what changed in the plant afterward.
• Ask which audit frameworks they supported and what evidence they personally gathered, since support in name only is common.
• Ask how they run change control when engineering wants to move fast. A good answer includes a fast path with limits, not no control.
• Ask what they know about liquid cooled racks, including what alarms matter and who responds at night.
• Ask how they plan capacity when a customer wants a cluster larger than the room can power and cool.
• Ask how they built or inherited a team, and who they promoted from within.
• Ask how they divide work between their own staff and outside contractors on a site that changes monthly.
• Ask how they brief executives after an incident and what they refuse to guess about.

Where do we find these candidates?

These managers come from colocation operations, enterprise data centers, high performance computing centers and cloud operations teams. Colocation people know customer contracts and incidents. Enterprise managers know discipline. High performance computing managers understand the cluster side. Cloud operations people know software defined thinking but are often light on plant.

• These managers come from colocation operations, enterprise data centers, high performance computing centers and cloud operations teams. Colocation people know customer contracts and incidents. Enterprise managers know discipline. High performance computing managers understand the cluster side. Cloud operations people know software defined thinking but are often light on plant.
• They move when a stable site offers no growth, when an owner stops investing or when they want to work with the newest hardware. Some are tired of slow approvals. Others want equity. Be honest about the pace and the risk of a young company, since stable candidates will ask about funding and customer concentration.

Which hiring mistakes should you avoid?

Do not hire a classic facilities manager and expect instant comfort with cluster customers who measure uptime by the job and not by the hour. Do not hire a startup operator with no audit record, then discover that a customer contract requires evidence you cannot produce.

• Hiring a classic facilities manager and expecting instant comfort with cluster customers who measure uptime by the job and not by the hour.
• Hiring a startup operator with no audit record, then discovering that a customer contract requires evidence you cannot produce.
• Leaving the retrofit plan to the manager alone, when engineering, supply and the customer all need to agree on the windows.

Your search in 3 steps

1. Map your customers and contracts

   We ask which customers run on which halls, what service levels you signed and which audits you must pass this year.

2. Choose the risk you accept

   We help you weigh process maturity against startup pace, and say which candidates sit where.

3. Test the bridge

   We ask candidates to explain a plant fault to a cluster engineer and a cluster fault to a plant engineer, and we note where it becomes vague.

Explore the relevant desks

• Data Center Operations Manager
• AI / GPU cloud providers (neoclouds)

Start your search

Related hiring resources

• Data Center Operations Manager recruiting
• AI / GPU cloud providers (neoclouds)
• Hiring guide: Critical Facilities Operations
• Every Critical Facilities Operations title
• How long a search takes
• Interview questions for Data Center Operations Manager
• MOPs, SOPs and EOPs: what candidates should have written or followed
• Switching authority and the qualified person, in hiring terms
• Shift rotations in critical operations and how they change the search