I build the platforms 100-engineer teams ship on — multi-account AWS, EKS at scale, GitOps from day one.

10+ years in cloud infrastructure. Currently running the platform at SumerSports — 10 EKS clusters, 15 AWS accounts, 100 engineers. Previously Senior SRE at Moonswitch, operating 12 EKS clusters across 5 client environments including Azure Government Cloud air-gapped workloads.

AWS Certified: DevOps Professional · SysOps Associate

Some things I've shipped:

• Reduced full-environment disaster recovery from 72 hours → 12–18 hours by codifying all client infrastructure in Terraform
• Led a 12-month program that recovered $600k from a $1.2M undocumented Azure footprint (approved budget: $0)
• Cut Aurora cluster provisioning from 10 business days → 30 minutes via self-service IaC bundles (Terramate Catalyst)
• Diagnosed cascading failures in a 65-node EKS cluster running 250+ preview environments — DNS quota exhaustion, 800+ leaked DB connections

Most of my production contributions live under my work account.

Work Account: @mthurman-sumer | Org: SumerSports

Python CLI for Kubernetes upgrade readiness assessment. Built to replace manual pre-upgrade checklists with scored, CI-gateable output.

• 0–100 readiness scoring engine with weighted categories across deprecations, addon versions, CRD compatibility, workload health, and security policies
• 12 typed exit codes for CI/CD pipeline integration — gate on specific failure conditions, not just pass/fail
• MCP server integration exposes checks as callable tools for AI agents
• Containerized deployment: Docker image with all required tools and MCP servers, designed to run as a scheduled scan
• v3.5.0 · 123+ merged PRs

A cloud-native personal AI memory system — capture thoughts from any Claude client, retrieve them semantically from anywhere.

MCP server running on Fly.io, storing knowledge in Supabase Postgres with pgvector. Every thought is embedded with Voyage AI (voyage-3-lite, 512-dim HNSW index) and linked into a typed knowledge graph you build yourself.

• 6 MCP tools: capture_thought, search_thoughts, list_recent, link_thoughts, get_links, stats
• Metadata extraction (people, topics, action items) via claude-haiku-4 through OpenRouter
• /capture REST endpoint for iOS Shortcuts — capture from any device, query from any Claude session
• Typed directional knowledge graph: related · supports · contradicts · follows_from · part_of · example_of · references

Infrastructure scaffolding framework: composable AWS patterns provisioned via a single catalyst scaffold command.

• Architected a 5-layer bundle system (Objects > Components > Bundles > Stacks > Generated Files) consumed across a 15-account AWS org
• Built RDS Aurora Global Cluster bundle: multi-region failover, IRSA integration, CloudWatch monitors, Datadog observability, KMS encryption
• Implemented typed schema validation, Checkov security scanning, and automated code generation
• Components include: IRSA roles, Route53 subdomains, ArgoCD applications, S3 buckets

Technologies: OpenTofu/Terraform · Terramate Catalyst · AWS (Aurora, EKS, KMS, Route53, ACM) · HCL

Multi-account AWS platform serving 9+ environments — OpenTofu, Terramate, and ArgoCD at the core.

• Managed EKS clusters with Karpenter autoscaling, node group normalization, and ALB controller integration
• Built Aurora Serverless v2 with cross-region failover and environment-specific capacity tuning
• Implemented Serverless PrivateLink with Lambda-automated RDS failover for cross-VPC database connectivity
• Provisioned CloudFront CDN, KMS key management, and SQS queues for data pipelines
• 1,293 commits · 100+ merged PRs · 11 repositories · trunk-based development

Technologies: OpenTofu · Terramate · AWS (EKS, Aurora, CloudFront, Lambda, PrivateLink, KMS, SQS) · Karpenter

Kubernetes application deployment platform across multiple EKS clusters, managed entirely through GitOps.

• Defined ArgoCD ApplicationSet patterns for multi-cluster system application deployment
• Configured Datadog agent for RDS monitoring and APM tracing
• Managed Helm chart lifecycle for C#/.NET and TypeScript microservices
• Led Kafka/Strimzi decommissioning across production clusters
• Deployed LiteLLM AI gateway with autoscaling and GitHub Actions Runner Controller (ARC)

Technologies: ArgoCD · Helm · Kubernetes · Datadog · GitHub Actions

Reusable infrastructure modules consumed across all stacks in a multi-account AWS organization.

• Aurora RDS module with KMS key rotation and environment-specific configuration
• Serverless PrivateLink module with Lambda-automated cross-VPC failover
• IRSA role module with configurable trust policies
• CloudFront module with S3 origin access and cache behavior ordering

Technologies: OpenTofu/Terraform · AWS

DevOps / Platform Engineer — SumerSports (2025 – Present) Own the full infrastructure lifecycle across a 15-account AWS org: IaC authoring, GitOps deployment, production monitoring. 1,293 commits and 100+ merged PRs across 11 repositories in year one.

Senior Site Reliability Engineer — Moonswitch (Prior) Operated 12 EKS clusters across 5 client environments. Reduced disaster recovery time from 72 hours to 12–18 hours by codifying all client infrastructure in Terraform. Led a $600k Azure cost recovery program. Managed Azure Government Cloud air-gapped workloads under FedRAMP/FISMA/NIST-800-53 compliance frameworks.