Founding Head of AI Infrastructure

Founding Head of AI Infrastructure Spacebus | Hire 01 | Full-time | Reports to: Founder | Regular site travel The opportunity Spacebus is at the 0–1 stage of building a distributed AI inference infrastructure business. Its proposed long-term architecture spans 10 sites, each with 14 NVIDIA Vera Rubin NVL72 racks (140 compute racks and 10,080 GPUs in total). This is a target architecture, not a representation that all sites, hardware, funding or customer contracts are committed. The immediate mandate is to validate and deliver the first commercially viable deployment, then establish a repeatable model for expansion. You are the first accountable technical leader: part architect, part builder, part technical co-creator of the business. You will turn ambition into a funded, sequenced, technically defensible roadmap, recruit the right specialists at the right time, and personally unblock critical technical work. This is not an inherited operations function. Accountabilities and deliverables • Own the technical roadmap from feasibility through first paid deployment and repeatable multi-site expansion. Define stage gates, dependencies, owners, critical path, evidence and contingency options; report progress and trade-offs to the Founder weekly. • Define the first sellable infrastructure offer with the Founder: dedicated capacity versus managed inference, workload profiles, tenancy, connectivity, performance baselines, security boundaries, customer onboarding and support obligations. Join customer discovery and technical sales; do not promise unvalidated capability. • Lead architecture and practical validation across compute, NVLink rack domains, inter-rack fabric, storage, management, provisioning, serving stack and telemetry. Select build/buy/partner approaches and prove key assumptions through labs, OEM tests or pilots before scale commitments. • Own the first-site technical work breakdown structure: facilities interfaces, power/liquid cooling dependencies, OEM and integrator responsibilities, procurement lead times, commissioning, acceptance testing, operational handover and customer readiness. Qualified specialists retain responsibility for regulated electrical/mechanical work. • Build the technical organization: map workstreams and skills, identify work that must be in-house versus outsourced, write role scorecards, propose hiring order and timing, interview and select candidates with Founder approval, onboard hires and establish accountability. Start lean; do not assume a 24/7 service can be supported by two people. • Create and manage an integrated technical budget: one-time hardware and integration costs, facilities interfaces, network/storage/software, contractor fees, staffing, travel, spares, support and ongoing operating costs. Produce base/downside scenarios, cash timing, vendor comparisons and cost per usable/sellable GPU-hour; maintain forecast versus actual. • Own technical risk, security, production readiness and reliability standards proportionate to each stage: acceptance evidence, configuration baselines, access controls, incident response, service-level objectives, recovery testing, OEM support and site-failure containment. • Be hands-on when needed: troubleshoot alongside the founding engineer, challenge vendors, review designs and tests, and close gaps rather than delegating an undefined problem. First 90 days: evidence-based outcomes Required outputs Days 1–30 Validate purchased/committed versus aspirational scope; map product/customer requirements; architecture options and key risks; first-site work breakdown and dependency map; draft hiring sequence; costed MVP budget and milestone roadmap. Days 31–60 Decide initial stack and vendors subject to approvals; demonstrate representative workload in available lab/partner environment; publish commissioning/acceptance plan; build hiring scorecards; initiate priority recruitment and procurement as funding allows. Days 61–90 Present benchmark and readiness evidence; lead first-site execution where equipment/facilities permit; support customer pilot or technical evaluation; publish operating/support model and revised multi-site expansion economics. Hardware-dependent commissioning dates are conditional on purchase, delivery, site readiness and approved spend. Success is judged on validated decisions and execution against agreed stage gates, not artificial deployment dates. How success is measured • First commercial deployment meets agreed customer and technical acceptance criteria; customer commitments are deliverable and documented. • Roadmap milestones, risk decisions and budget forecasts are transparent; variances and recovery plans are raised early. • Architecture can be repeated economically; cost, usable capacity, throughput and reliability are measured, not assumed. • Each workstream has a named accountable owner; hiring and outsourcing occur in the agreed sequence, within approved headcount and spend. What you bring • Demonstrated hands-on delivery of production GPU/HPC/cloud infrastructure, including Linux, networking, orchestration, storage, automation and troubleshooting; strong NVIDIA ecosystem experience is valuable. • Evidence of taking ambiguous, resource-constrained projects from concept to first production customer, not only operating mature fleets. • Practical understanding of dense GPU facilities dependencies, OEM/integrator coordination and acceptance evidence. • Ability to build a technical team and make capital-aware build/buy decisions; communicate clearly with founders, customers and investors. • Comfort doing detailed engineering work and making decisions with incomplete information, while explicitly tracking assumptions. Founder–hire working contract The Founder owns company strategy, financing, commercial commitments, capital allocation and final approval of headcount, compensation, material contracts and spending limits. The Head owns technical recommendations, execution plans, architecture within approved constraints, team design, technical hiring assessments and delivery of agreed milestones. Neither party commits the other to dates, SLAs, capacity or spending without explicit agreement. Budget envelope, delegated authority, compensation and initial site commitments must be agreed in writing before offer acceptance.