You do not need a platform team or a Kubernetes cluster to deploy safely. A small team can get fast, reliable deploys with GitHub Actions and a few AWS services. Here is a lean setup that scales with you.
What "good" looks like for a small team
- Every push runs tests automatically
- Merges to main deploy to staging without manual steps
- Production deploys are one click (or one approval)
- Rolling back is fast and obvious
The pipeline shape
Keep it boring. Three stages, test, build, deploy, triggered by git events. Use OIDC so GitHub Actions assumes an AWS role instead of you storing long-lived keys as secrets.
name: deploy
on:
push:
branches: [main]
permissions:
id-token: write # OIDC, no stored AWS keys
contents: read
jobs:
test:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- run: npm ci && npm test
deploy:
needs: test
runs-on: ubuntu-latest
steps:
- uses: aws-actions/configure-aws-credentials@v4
with:
role-to-assume: arn:aws:iam::123:role/deploy
aws-region: us-east-1
- run: ./deploy.shUse OIDC, not stored keys
The single most common CI/CD security mistake is pasting AWS access keys into repo secrets. GitHub Actions supports OIDC: it exchanges a short-lived token for a scoped AWS role. No long-lived credentials to leak.
If your CI has permanent AWS keys in its secrets, that is the first thing to fix, before anything about speed.
Make rollback trivial
Deploy immutable artifacts (a container image or versioned bundle) so rolling back is redeploying the previous version, not reverting code and rebuilding. On ECS or App Runner, that is pointing back at the last known-good image.
Do not over-build it
A small team does not need canary analysis, service meshes, or a bespoke internal platform on day one. Start with test → build → deploy plus easy rollback. Add complexity only when a real problem demands it.
If you want this set up cleanly the first time, OIDC, staged deploys, and a rollback you have actually tested, that is what our CI/CD work delivers.
About the author
Deep Mehta
Deep is the founder of 3 Dices Technology, a cloud engineering studio. He has shipped AWS architecture, DevOps automation, and production AI systems for startups and SMBs, and writes about the practical version of that work, not the conference-talk version.
Connect on LinkedIn