Systems Architect – AI Infra & GPU
Dubai / Abu Dhabi / Remote . Full Time | Flexible – AI Infrastructure & GPU Systems Architect with 10+ years and must have AI data center builds design experience.
Goal: Design the architecture. Build the systems. Prove the performance.
The Opportunity:
At DeFiTech, we are looking for an AI Infrastructure & GPU Systems Architect with the practical depth of an expert GPU Infrastructure Engineering. You will take ownership from workload assessment and architecture through deployment, optimization, and production readiness. This role requires strong foundations across AI compute, GPU memory, high-performance networking, storage, and liquid cooling, with the ability to make these components work together as a complete system.
Your scope will span enterprise AI platforms, Sovereign Infrastructure, GPU cloud environments, and AI data centers—supporting complex training, fine-tuning, inference, and high-performance computing workloads across different vendors and deployment scenarios.
What you will own:
- Turn workload requirements into infrastructure. Translate model size, data volume, concurrency, latency, and growth targets into practical architectures, capacity plans, and equipment specifications.
- Design and build GPU platforms. Deliver solutions ranging from individual GPU servers to distributed clusters across on-premises, colocation, cloud, and hybrid environments.
- Select technologies on merit. Evaluate GPU, server, network, storage, and cooling options against workload performance, compatibility, reliability, availability, and total cost of ownership.
- Lead implementation. Configure and integrate hardware, firmware, drivers, operating systems, GPU software, and orchestration platforms; resolve issues across the complete stack.
- Validate performance under real workloads. Establish acceptance benchmarks, identify bottlenecks, and tune systems for throughput, latency, utilization, stability, and energy efficiency.
- Deliver systems that teams can operate. Produce clear architecture diagrams, bills of materials, deployment guides, monitoring requirements, and operational handovers. Work closely with AI teams, vendors, and data center specialists.
The technical depth you bring
AI Compute & GPU Systems
Deep understanding of GPU architecture, accelerated computing, CPU–GPU balance, PCIe and NUMA topology, and GPU interconnects. Practical experience with multi-GPU and multi-node workloads, distributed training, inference serving, and parallelism strategies.
AI Memory
Strong knowledge of GPU memory capacity and bandwidth, high-bandwidth memory, host RAM, and data movement. Ability to size memory for model weights, activations, optimizer states, and inference KV caches—and assess trade-offs involving precision, quantization, sharding, and offloading.
AI Networking
Hands-on experience with high-performance Ethernet and InfiniBand, RDMA/RoCE, fabric topology, congestion management, and collective communication libraries such as NCCL or RCCL. Ability to diagnose communication bottlenecks and design suitable compute, storage, and management networks.
AI Storage & Data Paths
Experience designing high-throughput data pipelines using NVMe, parallel or distributed file systems, and object storage. Understand how data loading, metadata operations, checkpointing, and recovery affect GPU utilization and workload completion time.
Liquid Cooling & Data Center Integration
Practical experience integrating high-density GPU systems with direct-to-chip liquid cooling. Understand rack power and heat loads, cooling distribution units, coolant loops, flow and temperature requirements, leak detection, and redundancy. Coordinate commissioning and facility interfaces with mechanical, electrical, and cooling specialists.
Platform Engineering & Operations
Strong Linux administration and troubleshooting skills, with practical experience in containers, Kubernetes and/or Slurm, automation, and infrastructure monitoring. Build for secure access, resource isolation, fault recovery, and maintainable operations.
What we are looking for
- Proven senior-level experience architecting and personally implementing production GPU infrastructure for AI or HPC.
- Deep expertise in at least one major GPU ecosystem, such as NVIDIA CUDA or AMD ROCm, with demonstrated ability to evaluate and integrate alternatives.
- Evidence of solving difficult problems across compute, memory, networking, storage, and thermal constraints.
- Ability to explain technical decisions clearly, document trade-offs, and lead delivery across engineering teams and technology partners.
- A relevant engineering or computing degree, or equivalent practical experience.
Additional strengths: Deployments across multiple GPU vendors, GPU cloud or multi-tenant platforms, large-scale cluster commissioning, and measurable improvements in workload performance or operating cost.
Apply with evidence of what you have built
Send your CV to career@defitech.net and include a brief example of a GPU platform you designed or delivered: the workload, infrastructure scale, your contribution, the hardest constraint, and the result.
Apply for The Position: Systems Architect – AI Infra & GPU
Please send your CV along with cover letter to careers@defitech.net