Teddy A.

I build and run infrastructure that doesn't go down: national government platforms for 2M+ citizens, mobile banking for 1M+ users, and GPU Kubernetes for AI workloads. Top Rated, 5+ years, and I ship verified, documented, and on time. WHAT I DO • Kubernetes done properly: RKE2/EKS/k3s, Helm, Rancher, Istio mTLS, GitOps • AWS architecture and cost control: Terraform-first, OIDC-only CI, zero long-lived keys • CI/CD developers trust: GitHub Actions, GitLab CI, Jenkins (4-hour builds → 15 min) • On-prem → cloud and cloud → cloud migrations with phased rollback plans • Observability that pages before users notice: Prometheus, Grafana, Loki, Mimir, Alertmanager • Hybrid networking most DevOps engineers can't do: MPLS, BGP/OSPF, WireGuard, VPC design RECENT WORK OpenFn (integration platform for governments and NGOs, 2026–present) — Deploy and harden on-premises Lightning for government health programmes under NDA: a three-server HA/failover design, an active-active clustered deployment for a multi-country programme, CIS Docker Bench hardening, a reusable k6 load-testing and capacity-benchmark toolkit, and on-site training for government server admins. OpenG2P Ethiopia, 3 ministries (2025–2026) — Led platform engineering for the national social-protection system: RKE2 + Rancher, Istio mTLS, Keycloak SSO, WireGuard site-to-site, XCP-ng on RAID 1+0, full Prometheus/Grafana/Loki stack, plus a large ODK → OpenG2P data migration with validation pipelines and audit trails. WeTruck (logistics SaaS) — Designed and ran a fully Terraformed AWS serverless stack: CloudFront → WAF (deny-by-default) → ALB → Lambda/FastAPI, RDS, DynamoDB, SQS + DLQ, Amplify for three Next.js apps. Own production incidents end-to-end; wrote the runbooks the team still uses. Exponent.ch (Switzerland) — NVIDIA GPU RKE2 clusters for AI workloads (n8n, Langfuse, LibreChat, ClickHouse). Shared-stack Helm across 5+ environments, Teleport zero-trust access, Mimir observability: 60% MTTR reduction, 3 outages caught before impact. MetaPlux (Tech Lead) — GPU-aware rendering platform across AWS/Azure with MPLS traffic engineering, BGP/OSPF across hybrid sites, WireGuard overlay. Six Ethiopian banks — HA mobile banking and USSD at 99.95% uptime, MPLS branch WAN with QoS, ELK at 10 TB/day, Proxmox + AWS hybrid (~35% cost cut). STACK Kubernetes (RKE2, EKS, K3s), Docker, Helm, Rancher, Istio · Terraform, Terragrunt, Ansible · GitHub Actions, GitLab CI, Jenkins, ArgoCD · AWS (Lambda, EKS, ECS, CloudFront, WAF, RDS, DynamoDB, SQS, IAM/OIDC, VPC), Azure, OpenStack · Prometheus, Grafana, Loki, Mimir, OpenTelemetry, ELK · Keycloak, Vault, Teleport · Proxmox, XCP-ng, VMware · PostgreSQL, Redis, MongoDB · Python (FastAPI), Bash HOW I WORK I overcommunicate early, give realistic timelines, and treat testing as the deliverable. If something's out of scope or a bad idea, you'll hear it before you spend money. Every engagement ends with a runbook the next engineer can follow. Message me for: a Kubernetes or AWS build/fix, a CI/CD pipeline overhaul, a migration plan with a rollback story, GPU/AI infrastructure, or an observability stack that actually alerts.