Role Profile
Job Title: Data Centre Architect
Reports to: Head of Data Centre Development
Location: UK, hybrid
Radiant is redefining how AI infrastructure is built. We design and operate AI-native cloud platforms engineered for sovereignty, performance, and scale. Our infrastructure powers GPU-native workloads, multi-tenant control planes, and high-performance AI systems designed for the most demanding environments. We are not building a generic cloud. We are building purpose-built AI infrastructure — from powered land, to compute, to software. As we scale, we're building this organisation from the ground up. We're looking for people who'd rather build the playbook than inherit one — who are comfortable doing the work themselves as well as directing others, and who can adapt as fast as our strategy does. This is a startup model: attitude and adaptability matter more to us than years of experience or the size of team you've previously managed.
This is a senior individual contributor role with authority to sign off technical schedules, working with cluster and network architects, NVIDIA and other OEM engineers, colocation provider engineering and sales teams, and the internal Procurement, Legal, Finance and Site Operations teams.
Задачи (Responsibilities)
You will be the single technical authority between our GPU-as-a-service business and the colocation providers that host our NVIDIA clusters. You turn cluster architecture and OEM requirements into a precise, contractable statement of work (SOW) for the data centre white space, then hold the provider to it through design, installation and acceptance. We are growing from tens of megawatts to hundreds, on rack-scale liquid-cooled platforms at 130 kW per rack and rising to 200 kW and beyond. Mistakes in power, cooling or fibre specification show up as stranded capacity, delayed cluster delivery and disputes, so the role carries real commercial weight.
Technical negotiation with colocation providers
•Lead technical due diligence on candidate sites: power availability and energisation dates, redundancy topology, liquid-cooling capability, floor loading, rack logistics, fibre and meet-me access.
•Negotiate the technical schedules of the colocation agreement with commercial and legal: capacity and density commitments, power and cooling SLAs, temperature and flow envelopes, metering, expansion rights, change control and remedies.
•Challenge provider claims with evidence: single-line diagrams, P&IDs, load-bank and commissioning records, reference customers.
Specification & SOW
•Work with cluster architects and OEMs (NVIDIA first) to capture rack, power, cooling, network and layout requirements, and keep them current across product generations.
•Write and own the technical SOW for white-space fitout: containment and cooling interfaces, power distribution to rack, structured cabling and pathways, rack and POD layout, security zoning, labelling and records.
•Define clear scope boundaries and responsibilities between us, the colocation provider, OEMs and installation contractors, with no gaps and no overlaps.
•Maintain a requirements baseline with traceability from OEM document, to SOW clause, to acceptance test.
Design, installation & acceptance
•Review provider and contractor designs, shop drawings and method statements against the SOW; run design reviews and close findings.
•Define acceptance: factory and site tests, liquid flush and cleanliness, power quality, thermal performance at full load, fibre loss, burn-in, and the evidence pack needed for each gate.
•Attend key witness tests and sign off, or reject, completion of each phase before hardware is energised.
•Manage technical change requests and variations, and advise on cost and schedule impact.
Knowledge & standards
•Build reusable reference designs, SOW templates and a site-scoring framework so each new deployment starts from a proven baseline.
•Track the roadmap: higher rack densities, 800 VDC distribution, new cooling classes and optics, and tell the business early what it means for sites already contracted.
Deliverables You Own
•Site technical assessment: scored, evidence-based report per candidate site, comparable across providers, with risks and mitigations.
•Technical schedules to the colocation agreement: measurable commitments, tested in acceptance, with remedies agreed by legal.
•Cluster white-space SOW: complete enough that a contractor can price and build without asking us questions; traceable to OEM requirements.
•Reference designs and layouts: rack, row and POD layouts, power and cooling single-line and P&ID level interfaces, fibre pathways and densities.
•Acceptance test plan and evidence pack: pass and fail criteria set before work starts; signed results per gate.
•Requirements baseline and change log: one current version of the truth, with every variation recorded and costed.
•Risk and issues register: technical risks to schedule and capacity, with owners and dates.
Working Relationships
•Cluster and network architects: you translate their cluster design into facility requirements, and they rely on you to say what a site can and cannot carry.
•OEMs, mainly NVIDIA: you are the facility-side counterpart; you obtain, interpret and challenge site-requirement documents.
•Colocation providers: day-to-day technical counterpart through negotiation, build and acceptance.
•Procurement, legal and finance: you provide the technical substance of commercial terms and the impact of variations.
•Site operations and deployment teams: you hand over a clean specification and acceptance evidence, and take back lessons learned.
Требования (Requirements)
•8 or more years in data centre design, delivery or technical consulting, including at least 3 years on high-density (50 kW per rack or more) or liquid-cooled deployments.
•Direct experience writing or reviewing technical specifications and SOWs for white-space fitout, and managing contractors against them to acceptance.
•Strong working knowledge of data centre power: MV and LV distribution, UPS and generator topologies, redundancy models (N+1, 2N, distributed), busway and rack power, metering and power quality.
•Strong working knowledge of cooling: direct-to-chip liquid cooling, CDUs, facility and technology loops, warm-water and dry-cooler designs, fluid quality and flushing, leak detection and containment.
•Working knowledge of structured cabling and pathways for large GPU fabrics (single-mode optics, high fibre counts, InfiniBand and Ethernet physical layers), and of rack and POD layouts.
•Familiarity with NVIDIA rack-scale platforms (GB200 or GB300 class, or later) and their facility requirements, and the ability to read OEM site-requirement documents critically.
•Proven ability to run technical negotiations with colocation or construction counterparties, and to explain trade-offs to commercial, legal and executive audiences.
•Command of standards such as EN 50600 and the Uptime Tier framework, ASHRAE TC 9.9 liquid-cooling guidance, and commissioning practice (levels 1 to 5).
•Excellent written communication: you produce clear, unambiguous contract-grade documents.
•Experience at a hyperscaler, neocloud or colocation provider on the operator or customer side.
•Experience with 800 VDC or other DC distribution, large battery storage, or power smoothing for AI load swings.
•Chartered engineer status or equivalent; recognised data centre credentials (for example CDCP, CDFOM, Uptime ATD or AOS).
•Familiarity with BIM, digital-twin tools and requirements-management tools.
•Willingness to travel up to 30% to sites in the UK, Europe and the US.
Success Measures
First 90 days: Audit current colocation contracts and SOWs and list technical gaps; publish a requirements baseline for the current NVIDIA platform; take over technical lead on one live negotiation; agree the site-scoring framework with the cluster architects.
6 months: One complete SOW issued and priced with no material scope disputes; acceptance plan agreed with the provider before build starts; reusable SOW template adopted by the team.
12 months: Every new site contracted on the template with traceable requirements; clusters accepted on schedule with no facility-caused performance issues; fewer late variations and a measurable cut in time from site shortlist to signed technical schedule.
Qualities we look for
•Set the standard: every single day, you spot opportunities to constructively shake things up.
•Inspire the change: there's no blueprint for the future. You'll embrace challenges and change.
•You're real and you're true to yourself: we cherish and celebrate diversity so you'll feel right at home whoever you are and whoever you're talking to, you treat everyone the same.
•Get stuck in: you'd rather be building than overseeing. Our strategy will keep evolving, and so will you — attitude and adaptability matter more to us than a fixed playbook or a big team behind you.
Условия (Conditions)
•Employment type: Full-time, permanent.
•Travel: Up to 30%.
•Interview process: Talent Screening > Hiring Manager Screen > Technical Interview > Final with the Head of Data Centre Development.
