Infrastructure should be invisible
When infrastructure is done well, nobody thinks about it:
- code merged to main reaches production in minutes
- a bad deploy is rolled back with one command
- the team hears about problems from an alert, not an angry customer
- the cloud bill is boring and predictable
Most small teams aren’t there, not for lack of skill, but because nobody has had the time to set it up properly.
What I’ve run in production
For my own game-network product, I built and operate the full infrastructure:
- Relay servers in several data centers, deployed by GitHub Actions.
- The API on Railway, the web platform on Vercel, and update artifacts on Cloudflare R2.
- A signed release pipeline for desktop auto-updates, with the signing key kept out of CI logs and repositories.
The same principles apply to your stack, whatever it is.
Typical findings in an audit
- Secrets in the wrong places. Passwords in text files, API keys committed to repositories, SSH keys shared over chat. These get rotated and moved into a proper secrets manager.
- No tested backups. A backup that has never been restored is a hope, not a backup.
- Single points of failure. One server, one person with access, one undocumented script.
- Paying for idle capacity. Oversized databases, forgotten instances and storage nobody reads.
Each finding comes with a fix, and most get fixed during the engagement itself, not left in a report.
Engagement options
- One-off setup or audit. A fixed-price project that ends with everything automated, documented and handed over. This suits teams that want to own their infrastructure themselves.
- Ongoing retainer. A few hours a month for monitoring reviews, dependency and security updates, cost checks and help with new services. This suits teams without an in-house DevOps engineer who still want someone accountable.
Either way, everything lives in your accounts:
- your cloud provider
- your GitHub organisation
- your secrets manager
Nothing depends on my personal access. That’s a principle, not a feature. The point of good infrastructure is that your business keeps running whoever is on the team. Access is granted for the engagement and can be revoked at the end in minutes.
Security basics, done properly
Most infrastructure problems I see in small teams are not exotic attacks. They’re basics that were never set up:
- Shared admin accounts.
- Production databases reachable from the whole internet.
- Missing two-factor authentication.
- No record of who deployed what.
Every engagement includes fixing these fundamentals:
- least-privilege access
- private networking for databases
- MFA everywhere
- an audit trail of deployments and changes
None of this is glamorous, but it’s what keeps a small company out of the news.