Senior AI Infra Architect for GPU Clusters

Yotta; Hinoba-an; On-site; PHP 9,404,000 - 13,166,000; Full time

Job summary

Yotta Data Services is seeking a senior AI infrastructure architect to design, deploy, and operate large-scale GPU clusters for distributed training and inference. You will lead multi-node GPU systems, optimize OS/kernel configurations, and craft compute blueprints for diverse AI workloads.

The role requires deep hands-on experience with NVIDIA GPUs, InfiniBand/RDMA, and high-performance storage, with a strong focus on security, reliability, and observability across multi-tenant environments.

Qualifications

• Deep hands-on experience with GPU systems (NVIDIA).
• Strong background in InfiniBand/RDMA/RoCEv2 and high‑performance storage.
• Experience designing large-scale AI infrastructure with 100+ GPUs is a plus.

Responsibilities

• Design and deploy large-scale GPU clusters for distributed training and inference.
• Architect multi-node GPU systems with NVLink/NVSwitch and PCIe Gen5.
• Define OS/kernel/driver/runtime configs for AI workloads (CUDA/ROCm, NCCL, UCX, OFED).
• Develop compute blueprints for training, fine-tuning, retrieval, and batch inference.
• Ensure secure, scalable, high-availability AI compute clusters.
• Implement monitoring and performance optimization (Prometheus, Grafana, DCGM).
• Architect secure multi-tenant GPU environments with IAM/RBAC and encryption in transit.

Skills

GPU systems

InfiniBand/RDMA/RoCEv2

High-performance storage

Multi-node GPU architectures

AI frameworks familiarity

Kubernetes GPU environments

Terraform

Ansible

Education

B.E./B.Tech or equivalent

Tools

MAAS/iPXE/Metal³

NVMe/NPF fabrics

Job description

Yotta Data Services is seeking a senior AI infrastructure architect to design, deploy, and operate large-scale GPU clusters for distributed training and inference. You will lead multi-node GPU systems, optimize OS/kernel configurations, and craft compute blueprints for diverse AI workloads.

The role requires deep hands-on experience with NVIDIA GPUs, InfiniBand/RDMA, and high-performance storage, with a strong focus on security, reliability, and observability across multi-tenant environments.