About the role
- Own and operate Kubernetes-based ML platform on EKS — supporting LLM inference (KServe), distributed compute (Dask/Ray), and workflow orchestration (Apache Airflow).
- Partner with data science and ML teams to design, deploy, and scale ML workloads — including GPU scheduling, autoscaling, resource isolation, and SLO-driven reliability.
- Architect, deploy, and optimize Ray clusters on Kubernetes for distributed ML workloads — enabling scalable training, batch inference, and low-latency serving with efficient CPU/GPU utilization.
- Design, build, and maintain cloud-native infrastructure across AWS (primary), GCP, and Azure — using Kubernetes (EKS / GKE / AKS), Terraform, Helm, and ArgoCD.
- Drive GitOps adoption and platform standardization — define reusable infrastructure patterns, Helm charts, and deployment workflows used across all product teams.
- Manage Kubernetes platform operations — cluster lifecycle, Karpenter-based autoscaling, multi-tenancy, and workload isolation for data science and engineering teams.
- Implement and maintain service mesh (Istio) — mTLS enforcement, traffic policies, and observability for inter-service communication.
- Maintain and improve the internal developer platform (Backstage IDP) — enabling self-service environments, service catalog, and onboarding workflows for engineering teams.
- Build and maintain full-stack observability infrastructure — metrics (Prometheus / Mimir), logs (Loki), traces (Tempo), and dashboards (Grafana) integrated with OpenTelemetry instrumentation.
- Define SLIs, SLOs, and error budget policies for production ML and platform services; lead incident response and post-mortem reviews.
- Proactively identify reliability risks and drive engineering improvements to maintain 99.9%+ uptime targets.
- Implement Kubernetes cost attribution and chargeback using Kubecost / OpenCost — driving per-team visibility and FinOps decision-making for AI infrastructure.
- Continuously optimize cloud spend through workload right-sizing, spot/preemptible usage, and resource scheduling strategies.
- Manage AWS multi-account governance using Control Tower, SCPs, GuardDuty, and IAM Identity Center — ensuring security posture across all environments.
- Own OIDC identity and SSO infrastructure integrated across internal tooling — Backstage, Airflow, and platform services.
- Support compliance and audit processes — ISO 27001, CIS Benchmarks, Well-Architected Reviews, and VAPT assessments.
- 5+ years of hands-on platform, DevOps, or SRE experience in production environments.
- Strong Kubernetes expertise — cluster operations, Helm, RBAC, autoscaling (Karpenter / Cluster Autoscaler), multi-tenancy; EKS experience preferred.
- Infrastructure as Code — Terraform (advanced), Ansible; experience managing large, multi-environment IaC codebases.
- AWS expertise — EC2, EKS, S3, RDS, IAM, VPC, CloudWatch, Control Tower, GuardDuty; GCP or Azure exposure is a plus.
- GitOps & CI/CD — ArgoCD, Bitbucket Pipelines / Jenkins, GitOps workflows at team scale.
- Observability — hands-on with Prometheus, Grafana, and at least one of: Loki, Tempo, OpenTelemetry, Datadog, or ELK.
- Scripting & automation — Python and Bash for tooling, automation, and platform integrations.
- Strong understanding of networking, security, and cloud cost management in Kubernetes environments.
- Experience with ML serving infrastructure — KServe, vLLM, Ray Serve, or similar model serving frameworks.
- Experience with Apache Airflow, Dask, or other data/ML pipeline orchestration at scale.
- Familiarity with Backstage or similar internal developer platforms (IDP).
- Istio or Envoy service mesh experience.
- FinOps tooling — Kubecost, OpenCost, or cloud provider cost management tools.
- OIDC / identity provider experience (Zitadel, Keycloak, or similar).
- AWS Certified Solutions Architect or equivalent cloud certification.
- Exposure to geospatial data workloads or satellite imagery pipelines.
- Bachelor's degree in Computer Science, Information Technology, or a related engineering discipline.
- Real Production Scale: LLM inference, geospatial data pipelines, and multi-cloud Kubernetes — not toy projects.
- High Ownership: You architect systems end-to-end. No tickets-only culture, no hand-holding required.
- Meaningful Impact: Your infrastructure powers products used by governments and institutions across the developing world.
- Growth & Benefits: Learning allowances, broadband, medical insurance, best-in-class leave policy, and hybrid work from Bengaluru.