Joel Freeman.
Hi, I'm Joel.
I'm a senior engineer at Starboard, where I lead platform and infrastructure. Before that I built weather data pipelines at MetService, modernised AWS accounts for more than a dozen customers at AC3, and started out on a support desk in 2019.
More about meCloud
Seven years across GCP, AWS and Azure, from landing zones and least-privilege access to finding savings in the bill.
- GCP
- AWS
- Azure
- Landing zones
- IAM
Experience
7 years · GCP · AWS · Azure- StarboardSenior Engineer, Platform & Infrastructure Lead
First infrastructure hire. I built the platform, moved production onto it, and now lead the team of four that owns it.
2025 - now - MetServiceDevOps Platform Engineer
Built the multi-account AWS delivery pipeline and the tooling that proved the new weather data platform matched the old one.
2023 - 2025 - AC3DevOps Engineer
Led modernisation across more than a dozen customer AWS accounts with CDK, Terraform and CloudFormation.
2022 - 2023
What I do outside
Training for my second marathon. I also produce and DJ, with over 300,000 streams and plays on BBC Radio 1 and George FM.
Projects.
A few things I've designed and shipped, from infrastructure to apps.
One place for every agent session I have running across Claude Code and Codex, and every promise I made in Slack or a meeting.
- Electron
- TypeScript
- SQLite
- Claude Code
- Codex
whatA local daemon keeps everything in one SQLite archive on my machine and works out where each session is at.
A commercial product for music festivals that I co-founded. Atlas brings a festival's ticket sales, its app and what happens on site into one place. I'm the sole engineer, so I built the platform and the mobile app and run both in production.
- Next.js
- Prisma
- Postgres
- Expo
- React Native
- Vercel Workflows
AtlasAtlas keeps one record per event, artist and attendee, so each person is known across every event and the app: spend, returning, home city and music taste.
Atlas APIOne Next.js app on Vercel serves the back office and the REST API, keeps staff and attendee sign-in apart, and checks every caller before it reads any data.
I built the platform, wrote the service contract every service ships through, and moved all 38 production services onto it with zero downtime. Deploys went from hours or days to five minutes, self-service, with one-click rollback.
- GKE
- Argo CD
- Kargo
- Kyverno
- OPA
- Helm
- Envoy Gateway
- Terraform
- GCP PAM
Service contractAbout ten lines per service drive its deployment, secrets, network policy, gateway routes and alerts, with no platform ticket.
- Service contract drives
- Deployment
- Secrets from Secret Manager
- Network policy
- Gateway routes
- Alerts
Writing
Long-form write-ups of real work: the decisions, the numbers, and what I would do again.
Capabilities and components: a platform needs a specification, not just an implementation
Everything you can point at in a platform describes how it works today. Nothing says what it's supposed to do. This is the model I'm using to fix that: capabilities, components, and contracts you can test.