Position: Strategic Program Manager (Data Center/AI Factory)
Location: REMOTE (may require a little travel if needed)
Employment type: Contract-to-hire
Job Description:
Client has 17 AI Factory deployments starting between September 2026 and February 2027, across Roche in Amsterdam and Phoenix, Groq in Houston, and Sharon AI in Australia and New Zealand. Seven of them start in December. On our first program at this scale, we did not have enough experienced client people on site at kickoff, and it cost us time and rework. The fix, approved by our EVP of Technical Delivery & Engineering, is a Tiger Team: one person each from PMO, Architecture and Field Services, plus a Logistics lead from Supply Chain. They land about a week before a build starts, get governance and reporting running, train the site crew by hand against our standard curriculum, confirm the crew absorbed it, and hand off to the permanent delivery team around two weeks in. Then they move to the next site. The permanent project team is on the ground with them from day one, so the customer builds its relationship with the people who are staying.
What the role does
This is the PMO seat on the Tiger Team. You own governance and reporting for the first two weeks of every new AI Factory deployment. You arrive before build starts, get the program controls running before anyone touches hardware, teach the PMO part of the curriculum to the incoming project team, and hand a working governance structure to the permanent program manager at the two-week mark. It is a working role, not an oversight role. We want someone who has personally run the messy first weeks of large on-site builds, has trained project staff by doing the work next to them, and can set expectations with senior PMs, customer stakeholders, and partner leads on day one.
Key responsibilities
Before the build starts
- Stand up the governance and reporting structure before the build starts: daily reporting, gate evidence, RAID and decision logs, escalation paths, and the tracking that feeds the AI Factory tool.
- Roll out the client's standard SOPs and receiving standards the same way at every site. Nothing gets improvised locally.
- Open the DOA, RMA, and expedite registers with the Logistics lead so material problems are visible and worked from day one.
Qualifications (Required):
- 8+ years in program or project management, with at least 5 years leading on-site delivery of large infrastructure, data center, or field deployment programs.
- Personally led at least three sites from kickoff through hand-off, including the first three weeks, with mixed crews of employees, contractors, partner labor, and customer staff.
- Has trained and ramped project staff by hand in a live delivery environment. Writing the documentation does not count.
- Disciplined about governance and evidence: receiving standards, SOP rollout, DOA and RMA handling, expedites, daily reporting, gate readiness.
- The ability to direct senior program managers and set expectations with customer and partner leadership on day one.
- Willing to travel internationally on short notice for up to three weeks at a stretch, and to be on a site in December.
Qualifications (Preferred)
- NVIDIA AI Factory, DGX or NVL rack deployment experience, or familiarity with the NVIS curriculum.
- Background in hyperscale or colocation data center build-outs, HPC cluster deployments, large multisite refresh programs, or field deployment programs with heavy logistics.
- Delivery experience in ANZ or EMEA. PMP or equivalent. Multisite bootcamp graduate or instructor.
- Has been the first person into a new site or a new program.
Requirements:
These terms represent the absolute cutting edge of enterprise AI data center architecture. Here is a breakdown of what each of these core components means in a deployment context:
An AI Factory is NVIDIA's architectural blueprint for a dedicated data center infrastructure designed solely to ingest data and produce AI models (like LLMs or computer vision systems). Unlike traditional cloud data centers that run mixed workloads (web servers, databases), an AI Factory treats compute, storage, and high-speed networking as a single, massive supercomputer optimized for parallel processing.
DGX vs. NVL Rack Deployments
Deploying these architectures requires deep expertise in high-density power (often 30kW to 100kW+ per rack), liquid cooling (facility water loops, CDU manifolds), and specialized optical networking.
- NVIDIA DGX Systems: These are standalone, ultra-high-performance AI servers (like the DGX H100 or B200). "DGX deployment experience" means knowing how to rack, stack, power, and cluster these individual 8-GPU nodes using PCIe and NVLink switches.
- NVIDIA NVL (NVLink Layer) Racks: This refers to fully integrated, rack-scale solutions like the GB200 NVL72. Instead of isolated servers, the entire rack acts as a single, giant GPU. Deploying NVL racks involves complex liquid-cooling connections, blind-mate infrastructure, and massive NVLink backplanes that interconnect dozens of Grace CPUs and Blackwell GPUs.
NVIS Curriculum
NVIS stands for NVIDIA Installation Services (or closely ties to the NVIDIA Certified Specialist certifications for cluster installation). Familiarity with this curriculum means you understand NVIDIA's official, strict standards for:
- Physical site readiness (floor load limits, liquid cooling requirements).
- Hardware validation and precision cabling guidelines.
- Running InfiniBand/RoCE fabric diagnostics.
- Deploying the NVIDIA Base Command Manager to initialize, test, and benchmark the cluster before handing it over to data scientists
