Job Specification: Solution Architect - NVIDIA Cluster (End-to-End Design & Validation)
Location: London (1 day per week onsite)
Travel: Occasional travel to datacenter sites outside the UK
Engagement: Contract Inside IR35
Department: Engineering/Advanced Compute
Role Overview
We are seeking a highly skilled Solution Architect with deep experience in designing, validating, and delivering end-to-end NVIDIA GPU clusters in enterprise and hyperscale environments. This individual will own the full life cycle of architectural design-from requirements gathering through implementation oversight and performance validation. They will work closely with engineering, networking, DevOps, security, and datacenter operations teams to ensure high-performance, scalable, and resilient GPU infrastructure for AI, HPC, and ML workloads.
The role is primarily London-based one day per week, with occasional international travel required to support datacenter design reviews, deployment validation, or site acceptance testing.
Key Responsibilities Architecture & Design
Lead the architecture of NVIDIA GPU clusters leveraging technologies such as H100/H200, NVLink, NVSwitch, DGX, HGX, or SuperPod-class designs.
Produce high-level and low-level designs (HLD/LLD), including compute, network, storage, and power/cooling considerations.
Validate hardware and platform selections, ensuring architectural alignment with customer requirements and scalability goals.
Design fabric architectures including InfiniBand (200/400Gb), RoCE, and high-performance east-west traffic patterns.
Ensure designs adhere to NVIDIA reference architectures (NVAIE, Base Command, DGX SuperPod specs, etc.).
Cluster Integration & Validation
Define and execute validation test plans for GPU cluster performance, resilience, networking throughput, and workload behaviour.
Oversee integration of GPU nodes, networking, and storage systems into the existing datacenter environment.
Collaborate with DevOps/Platform teams to validate cluster orchestration (Kubernetes, Slurm, Bright Cluster Manager, or equivalents).
Validate firmware, drivers, NCCL, CUDA libraries, and container environments for production readiness.
Deployment & Delivery Oversight
Provide technical leadership across the full deployment life cycle.
Partner with datacenter operations to ensure correct rack layouts, cabling, airflow and power design.
Support delivery teams during build-out phases, ensuring the design is executed correctly.
Participate in factory acceptance tests (FAT), site acceptance tests (SAT), and operational readiness reviews.
Stakeholder Collaboration
Work closely with internal and external teams including network engineering, platform engineering, procurement, and vendors such as NVIDIA, Mellanox, Supermicro, Dell, or HPE.
Provide technical guidance to customers, partners, and cross-functional engineering teams.
Communicate complex architectural concepts clearly to both technical and non-technical audiences.
Documentation & Governance
Produce detailed architecture documents, diagrams, acceptance criteria, and operational runbooks.
Ensure security, compliance, and governance standards are built into the design.
Provide knowledge transfer (KT) and training sessions to internal teams where required.
Required Skills & Experience Technical Expertise
Proven experience architecting and delivering NVIDIA GPU clusters at scale (AI/ML/HPC environments).
Strong hands-on understanding of GPU interconnects (NVLink/NVSwitch) and DGX/HGX/SuperPod architectures.
Deep knowledge of InfiniBand and high-performance networking architectures.
Experience with cluster orchestration: Kubernetes, Slurm, PBS, or similar.
Familiarity with AI/ML workload requirements, CUDA, Docker/OCI containers, and NVIDIA software stacks (NCCL, CUDA Toolkit).
Comfort with Linux systems engineering, hardware validation, and troubleshooting across compute/network layers.
Soft Skills
Strong communication skills, with the ability to bridge engineering and business discussions.
Comfortable owning architecture decisions and delivering executive-ready documentation.
Ability to work autonomously while coordinating with multi-disciplinary teams.
Problem-solver with strong critical-thinking abilities and a delivery-focused mindset.
Desirable Experience
Experience with hyperscaler-class deployments or multi-megawatt datacenter environments.
Work with NVIDIA Base Command Manager or similar cluster management tooling.
Exposure to data pipelines, storage systems (Lustre, GPUDirect Storage, Ceph), or AI workflow platforms.
Certifications such as NVIDIA Certified Associate/Expert, Kubernetes certifications (CKA/CKS), or related vendor accreditations.
What We Offer
Hybrid working: 1 day per week in London
Opportunity to design next-generation high-performance GPU infrastructure
Exposure to cutting-edge AI compute at scale