Vikram S.

I keep AWS, Azure and Kubernetes platforms running for teams: CI/CD, Terraform, monitoring and incident response. On production Kubernetes I cut incident recovery time from about 45 minutes to under 10 minutes with Prometheus, Grafana, Loki and Tempo, and I reduced Azure spend by about 30%. I am a Top Rated Upwork freelancer working in DevOps, cloud, Kubernetes and platform engineering. I have led an 8-engineer DevOps team on AWS, scaled an e-commerce platform to about 4M daily visitors, and removed about $48K a year of wasted AWS compute. My focus now is AI platform engineering and LLMOps: turning AI demos, model-serving APIs and RAG systems into production services your team can operate. PROBLEMS I SOLVE Kubernetes or cloud infrastructure is slowing the team down - Diagnose cluster failures, 503 errors, pod crashes, ingress problems, CI/CD bottlenecks and unreliable deployments. - Build Terraform modules, GitOps with Argo CD, and CI/CD pipelines in GitHub Actions, GitLab CI, Jenkins or Azure DevOps. - Improve security with RBAC, network controls, secrets management, image scanning and least-privilege cloud access. Cloud bills keep growing - Cost reviews, rightsizing, reserved and Spot capacity, and cost attribution on AWS and Azure. AI app works in a demo but is not production-ready - Deploy LLM and AI backend services with Docker, FastAPI, Kubernetes, ingress, HTTPS, secrets, logging and restart/recovery behavior. - Validate NVIDIA GPU, CUDA, drivers and container runtime, and check the workload really uses the GPU. - Model serving with vLLM and KServe on AKS/EKS GPU node pools, with metrics, alerts and runbooks. RECENT RELEVANT WORK - Production Kubernetes on AKS, EKS and DOKS for fintech, e-commerce and SaaS clients, moved to GitOps with Argo CD ApplicationSets. - Observability stack with service-level alerting: incident recovery from about 45 minutes to under 10 minutes. - Azure spend down by about 30% through rightsizing, cost reviews and Spot-capacity planning. - Led an 8-engineer DevOps team: reusable Terraform and EKS modules across accounts, blue-green deployments after a production outage. - Automated SOC 2 Type II identity, logging and change-management controls for a fintech client; Trivy image scanning in CI. - Production AWS/EKS support and incident handling for a pharma client's engineering teams across time zones. - Llama 3 8B on AKS GPU node pools with vLLM, KServe, scale-to-zero and Spot capacity: about 504 tokens/sec and about $0.29 per 1M output tokens. WHAT YOU CAN HIRE ME FOR - Kubernetes platform build, rescue or review (AKS, EKS) - Terraform infrastructure automation - CI/CD and GitOps (Argo CD, GitHub Actions, GitLab CI, Jenkins, Azure DevOps) - Prometheus/Grafana observability and incident response - Cloud cost and reliability review - Production LLM deployment and GPU setup (vLLM, KServe) IMPORTANT FIT I am not an AI researcher, data scientist or model-training specialist. I make the platform under the application reliable, secure, observable and cost-aware. If you are dealing with Kubernetes instability, slow CI/CD, cloud cost surprises or an AI feature that needs to become a production service, send me a short description of your setup. I will tell you what I can fix, what I would check first, and a realistic delivery plan.