DevOps engineer (India) planning to move into AI/GPU infrastructure. Is this the right path and worth it?
retained text
Hi everyone, I'm a DevOps engineer with 2.5 years of experience, based in North India. My current work: EKS clusters (dev/staging/prod) running 20+ microservices GitOps with ArgoCD, Terraform, GitLab CI Observability with Prometheus, Grafana, Loki, OpenTelemetry Custom Kubernetes operators written in Go I want to move into a role that stays relevant as AI automates more routine DevOps work. After some research, I'm leaning towards AI/GPU infrastructure: running GPU clusters for training and inference. India's GPU and data centre capacity is growing fast (IndiaAI Mission, Yotta, E2E, Neysa, etc.), and I'm seeing roles at NVIDIA, Deloitte, HCLTech and GPU cloud startups. My learning plan (3–4 months): GPU fundamentals: drivers/CUDA compatibility, NVLink, nvidia-smi topology GPU on Kubernetes: GPU Operator, MIG/MPS/time-slicing, DCGM monitoring, Kueue Distributed training ops: PyTorch DDP, NCCL tests and debugging Networking: RDMA, InfiniBand vs RoCE, GPUDirect Slurm basics, Lustre concepts, Linux tuning (NUMA, hugepages) Portfolio project: Rent 2 GPU nodes for a few hours, set up K8s + GPU Operator via GitOps, run multi-node NCCL tests, monitor with DCGM + Grafana, and write up what broke. Certs: CKA (if needed), NVIDIA NCA-AIIO first, NCP-AII later. My questions: Is this a realistic switch from DevOps, or do companies only hire people with production HPC/InfiniBand experience? For those already in GPU/HPC infra in India: is the demand real, or mostly hype around announcements? Is K8s-based GPU infra the better entry point, or should I focus on Slurm/traditional HPC? Is a lab project with rented GPUs enough to get shortlisted, or is it worth taking a lower role just to get real cluster exposure? How does pay compare with senior DevOps/SRE roles? Anything important missing from my plan? Would really appreciate honest views, especially from people working at NVIDIA, GPU clouds, or HPC centres. Thanks! TL;DR: DevOps engineer with K8s/Go/GitOps experience, planning a 3–4 month switch to AI/GPU infrastructure in India. Is it worth it, and is my plan realistic?
document fetched 2026-10-15 00:25 · canonical https://reddit.com/r/developersIndia/comments/1x5xw0g/devops_engineer_india_planning_to_move_into_aigpu