Role: DevOps Engineer

Function: DevOps / Platform Engineering / SRE

Location: Mumbai / Bangalore

Type: Full-time

Compensation: Not specified

Industry: Consumer AI

About Company

The company is building the AI layer for Bharat at India-scale. It is backed by partnerships with global tech leaders including Meta and Google.

The mission: AI that serves the entire Indian user base — across languages, contexts, and daily needs. Engineered from day one for 100M+ users and 1B-ready constraints on latency, cost, reliability, and safety.

The culture runs like a startup inside a high-scale ecosystem — small pods, high autonomy, and fast iteration. Engineering excellence is non-negotiable here.

Position Overview

This role owns the infrastructure backbone that every engineer, model, and product at the company depends on — from multi-region GCP infrastructure and multi-cluster Kubernetes environments to GPU compute for ML workloads and the CI/CD pipelines that ship at high velocity. The successful candidate will build 0-to-1 platform infrastructure designed from day one for extreme scale, low latency, and high reliability, while supporting the specialised compute demands of real-time AI inference and training pipelines serving hundreds of millions of Indian users.

Role & Responsibilities

• Design, provision, and operate production-grade multi-region GCP infrastructure — owning the full stack from VPCs, IAM, and networking to compute, storage, and cloud security posture management
• Architect and operate multi-cluster GKE environments: cluster lifecycle management, autoscaling (HPA, VPA, KEDA, Cluster Autoscaler), workload scheduling, network policies, and upgrade strategies
• Build and maintain high-velocity CI/CD pipelines with blue-green, canary, and rolling deployment strategies; own GitOps workflows via ArgoCD or Flux for declarative, auditable Kubernetes delivery
• Provision, schedule, and optimise GPU node pools for ML training, fine-tuning, and real-time inference — including MIG/time-sharing configurations, spot/preemptible handling with checkpointing, and cost-aware autoscaling
• Build and own the full observability stack — metrics (Prometheus, Cloud Monitoring), structured logs, distributed traces (OpenTelemetry), and SLI/SLO dashboards — and lead end-to-end incident response including PIR documentation and preventive automation
• Implement and maintain full IaC coverage using Terraform at module/workspace scale — including state backend management, drift detection, and automated plan/apply pipelines
• Own cloud cost governance: resource tagging, utilisation monitoring, rightsizing, commitment-based discount strategies, and automated waste detection across GPU and general compute

Must Have Criteria

• 5–8 years in DevOps, SRE, or platform engineering with hands-on ownership of production systems at scale — not tooling maintenance or pipeline-only roles
• Deep GCP expertise (mandatory): GKE, Cloud Run, GCS, VPC, IAM, Secret Manager, Cloud Monitoring, Pub/Sub, and Artifact Registry in production environments
• Production-grade Kubernetes: multi-cluster lifecycle management, autoscaling architectures (HPA/VPA/KEDA/Cluster Autoscaler), RBAC, network policies, and workload isolation
• Terraform ownership at module/workspace scale: state backend management, drift detection, and automated apply pipelines in a multi-engineer team environment
• CI/CD pipeline design and ownership for multi-service, high-velocity environments including GitOps delivery with ArgoCD or Flux
• Hands-on GPU compute provisioning for ML workloads: NVIDIA node pools, device plugin configuration, MIG or time-sharing strategies, spot/preemptible handling, and cost-aware autoscaling
• Observability engineering from scratch: Prometheus, Grafana, structured logging, distributed tracing (OpenTelemetry), alerting policy design, and production incident response with RCA documentation and follow-through
• Working proficiency in at least one scripting/programming language — Python (preferred), Node.js, or Java — for automation, tooling, and infrastructure utilities; this is not primarily a coding role but technical depth matters

Nice to Have

• AI serving infrastructure experience: Triton Inference Server, vLLM, or Ray Serve — including model versioning and A/B traffic routing
• Service mesh implementation: Istio or Cilium for mTLS, traffic management, and zero-trust networking in production
• FinOps tooling: Kubecost, Infracost, or equivalent for granular GPU and cloud cost attribution and rightsizing
• Container security: image scanning (Trivy/Snyk), admission controllers (OPA Gatekeeper/Kyverno), and runtime security (Falco)
• Chaos engineering facilitation: Litmus, Chaos Monkey, or structured game-day exercises for platform resilience validation

What We Offer

• Foundational infrastructure ownership at India-scale — systems built for 100M+ users from day one, with a clear path to 1B-ready architecture
• Direct exposure to cutting-edge AI compute infrastructure: GPU clusters, real-time inference serving, and multimodal ML pipelines
• Startup-pace execution within the company's high-scale ecosystem — small pod, high autonomy, fast iteration, and outsized impact on a category-defining consumer AI product
• Collaboration with ML engineers, platform architects, and product leadership on systems that matter to hundreds of millions of Indian users