Ehtisham M.

I help SaaS and AI teams build, migrate and stabilize production Kubernetes on AWS. Platform engineering, EKS, Terraform, CI/CD, cloud migration, cost control, and the GPU and inference infrastructure AI products need. CKA and CKAD, hands-on. Most of my work falls into one of five problems: 1. AI that works on a laptop but not in production. AI agents, LLM and RAG systems that fall over under real traffic. Evals, guardrails, GPU scheduling, inference cost control, CI/CD for models. This is where most AI projects actually die. 2. Kubernetes that keeps breaking. Nodes evicting, deploys that need a human at 2am, no observability. I make clusters boring. 3. Infrastructure you inherited and nobody documented. The original builders are gone, nothing is in code, and you are afraid to touch it. I take it over, map it, get it into Terraform, and hand you back something you can actually change. 4. Cloud bills that grew faster than the product. I cut a Danish SaaS company's spend by 40 percent with Terraform and multi-cloud rightsizing, and held a 99.9 percent SLA through the change. 5. Migration. AWS account to AWS account, on-prem to cloud, cloud to cloud, VM to Kubernetes. Planned, staged, and reversible. Selected work: - Built a multi-tenant Kubernetes platform for a funded UK company: 15+ cloud providers, GPU compute, a 100+ component marketplace. Live in production. - Designed Kubeflow and multi-cloud Kubernetes for a Norwegian ML platform. Still running in production 5+ years later, GDPR compliant. - Ran an on-prem Kubernetes and GPU research cluster for Imperial College London. 100 percent uptime across 2 years, bare metal. - Hardened Kubernetes and IaC for a Saudi fintech under SAMA requirements. - Architected HIPAA-aligned AWS infrastructure for a US healthcare SaaS. Stack: Kubernetes (EKS, AKS, GKE, K3s, RKE2, Rancher), Terraform, ArgoCD, GitOps, GitHub Actions, Helm, Docker, Linux, Istio, Vault, Keycloak, Prometheus, Grafana, Kubeflow, MLflow, PostgreSQL high availability, self-hosted and multi-tenant platforms, n8n, AWS, Azure, GCP, Python, FastAPI. SRE and platform engineering. Compliance: HIPAA, GDPR, SOC 2, SAMA. I work 30+ hours a week with 4 to 6 hours of overlap with US business hours, and I reply within a few hours on weekdays. Two ways to start. A fixed-price Kubernetes and cloud audit: I review your architecture, cost and reliability posture and hand back a prioritized remediation plan you own whether or not we work together. Or embed part-time on an ongoing basis, from 5 to 10 hours a week upward, as your platform engineer. Send me your current architecture or the specific failure you are hitting and I will tell you how I would approach it.