Director of Technical Program Management, Logical Deployment

Description

The company builds and runs large-scale AI infrastructure, supplying GPU clusters to hyperscalers, AI labs and enterprises in North America, EMEA and APAC.

The Deployment PMO takes signed contracts and gets them to live AI infrastructure, coordinating the many teams and partners that bring large GPU clusters online.

This Director role runs logical deployment. Once racks have power, you lead the network and software bring-up that turns them into a working, validated cluster ready for the customer.

Your scope covers asset and inventory data, bootstrap connectivity, fabric deployment, control plane and fleet bring-up, GPU enrollment, acceptance validation and handoff to operations.

Your job is to turn a tangle of engineering dependencies into a schedule the business can rely on, and to keep engineering, networking and partner teams accountable for it.

Success in your first 6 to 12 months: each active cluster has one integrated logical schedule with a named lead, reviewed weekly. Logical milestones hit their dates, or slips are flagged at least 2 weeks ahead with a recovery plan. Time from rack energization to handoff gets shorter with each cluster. Acceptance validation follows a documented, repeatable standard on every site. Clusters move to operations with no open blockers. A written logical deployment playbook and core bring-up automation are used on every new site. The team is fully staffed, with clear ownership and a succession path.

Responsibilities

• Own logical bring-up.
• Own one integrated logical schedule per cluster, running from asset data lock through to customer handoff.
• Keep asset and inventory data in NetBox as the source of truth. Lock it before bring-up starts and keep it accurate throughout.
• Drive connectivity readiness: edge connectivity, out-of-band and management networks, and bootstrap connectivity for both company and customer environments.
• Sequence service, control plane and fleet bring-up across engineering teams.
• Drive front-end and back-end fabric deployment to completion with the Networking and Build Out teams.
• Clear fabric, configuration and connectivity blockers fast, and escalate early when they threaten the schedule.
• Own GPU enrollment and validation, plus fleet manager ingestion and provisioning.
• Lead acceptance validation against customer criteria, and close defects with engineering and OEM partners.
• Prepare each cluster for handoff to ongoing operations, with complete documentation, runbooks and open-issue lists.
• Lead, hire and develop engineering program managers, networking and validation specialists, and tooling and automation leads.
• Write the logical deployment playbook, covering entry criteria from Physical Deployment, gates, checklists and standard timelines.
• Automate bring-up, validation and status reporting with the tooling team so each cluster comes online faster than the last.
• Report weekly status and risks to the VP and senior leadership. Separate forecast from upside clearly.

Requirements

• 12+ years in hyperscale infrastructure, cloud platforms or large-scale HPC delivery, including 5+ years leading teams.
• Direct experience delivering large GPU or HPC clusters, from network and platform bring-up through validation and customer acceptance.
• Solid working knowledge of data center networking: front-end and back-end fabrics, InfiniBand or RoCE, and out-of-band and management networks.
• Solid working knowledge of infrastructure software, including provisioning, control planes, fleet management and asset systems such as NetBox.
• A record of driving engineering teams you do not manage directly to firm commitments.
• Skill in building integrated schedules and critical paths across many dependencies, and recovering them when they slip.
• Clear, concise executive communication, in writing and in the room, with honest forecasts.
• Ability to travel about 50% of the time, mostly to data center sites.
• You make the real call on the date, even when it is unpopular, and pair it with a recovery plan. You decide on incomplete information and stand behind the decision.
• You hold engineering, networking and partner teams to their commitments. When a dependency slips, you escalate early, directly and with evidence.
• You own the outcome for every cluster in scope and explain what you are doing about problems rather than waiting on others.
• You steady a failing bring-up and help your team think clearly when schedules collapse.
• You brief SVPs and customers with a clear, honest position: what is true, what is at risk, and what you need from them.
• You hire strong engineers and program managers, give them real ownership, and develop them into leaders.
• You can read a fabric design, a bring-up log or a validation report and find the problem yourself. Status reports alone are not enough.
• Preferred: hands-on experience with NVIDIA GB200 or GB300 or Vera Rubin NVL72-class clusters, including NVLink, InfiniBand or Spectrum-X fabrics.
• Preferred: experience at a hyperscaler, neocloud or AI lab bringing 10K+ GPU clusters into production.
• Preferred: familiarity with NVIDIA cluster validation and burn-in.
• Preferred: experience building automation for provisioning, validation or deployment reporting.
• Preferred: a background in network engineering, SRE or infrastructure software.

Conditions

Base salary range: $290,000 - $350,000 USD.

Actual pay may vary based on job-related factors such as skill set, experience, education and location.

Travel is about 50% of the time, mostly to data center sites.

Benefits

• The role may be eligible for bonus, equity and/or commission programs.
• The company may offer a competitive benefits package with medical, dental and vision cover, flexible paid time off, parental leave and retirement plan participation.