AI Compute Engineer

DAMAC Digital; Bengaluru; On-site; INR 4,000,000 - 6,000,000; Full time

Job summary

DAMAC Digital in Bengaluru is seeking an experienced HPC engineer to design and optimize GPU compute clusters for training and inference workloads. You will own architecture, deployment, and performance across bare metal, firmware, CUDA, NCCL, cuDNN, and the orchestration layer.

You'll integrate InfiniBand/RoCE fabrics, Slurm, Kubernetes and Run:AI, benchmark with NCCL tests and MLPerf, and lead vendor engagements with NVIDIA and partners.

Qualifications

• 7+ years in HPC, AI or GPU infrastructure.
• 3+ years hands-on with NVIDIA GPU clusters.
• Deep NVIDIA stack expertise: CUDA, NCCL, cuDNN, TensorRT, DCGM.
• InfiniBand/RoCE networking knowledge and Slurm/Kubernetes/Run:ai experience.
• Advanced Linux, Python/Bash, Ansible or Terraform.

Responsibilities

• Architect GPU compute solutions for large-scale training and inference clusters.
• Deploy and tune NVIDIA Base Command Manager, CUDA, NCCL, cuDNN, TensorRT, DCGM, MIG.
• Integrate with InfiniBand/RoCE fabrics and onboard Slurm, Kubernetes, Run:ai.
• Benchmark with NCCL tests, MLPerf and HPL, and chase down bottlenecks.
• Own firmware/driver upgrade cycles and lead vendor engagements with NVIDIA, Supermicro, Dell, HPE.

Skills

NVIDIA GPU clusters

CUDA / NCCL / cuDNN

Networking InfiniBand

Slurm / Kubernetes

Run:ai

Python / Bash

Ansible

Terraform

Advanced Linux

Vendor management

Tools

NVIDIA Base Command Manager

DCGM

MIG

MLPerf benchmarking

Job description

Someone has to design and tune the GPU clusters that actually run DAMAC AI's training and inference workloads. That's this role.

We're building one of the region's most ambitious AI compute footprints — NVIDIA B200, B300 and GB300 NVL72 clusters powering sovereign cloud and hyperscale AI services across the Middle East and Asia. You'll own architecture, deployment and performance across the full stack: bare metal, firmware, CUDA, NCCL, orchestration and multi-tenant scheduling.

What you'll do

• Architect GPU compute solutions for large-scale training and inference clusters
• Deploy and tune NVIDIA Base Command Manager, CUDA, NCCL, cuDNN, TensorRT, DCGM, MIG
• Integrate with InfiniBand/RoCE fabrics and onboard Slurm, Kubernetes, Run:ai
• Benchmark with NCCL tests, MLPerf and HPL, and chase down every bottleneck
• Own firmware/driver upgrade cycles and lead vendor engagements with NVIDIA, Supermicro, Dell, HPE

What you bring

• 7+ years in HPC, AI or GPU infrastructure, 3+ years hands-on with NVIDIA GPU clusters
• Deep NVIDIA stack expertise: CUDA, NCCL, cuDNN, TensorRT, DCGM
• InfiniBand/RoCE networking knowledge and Slurm/Kubernetes/Run:ai experience
• Advanced Linux, Python/Bash, Ansible or Terraform
• Bonus: NVIDIA certification, LLM training at scale, liquid-cooled GB300 NVL72 experience

If you'd rather be tuning NCCL collectives than sitting in another status meeting — this is your seat.