CAPABILITY / 03

DevOps & release engineering

In production

Release paths that fail safely: tests gate the build, probes check the live system, and the previous images stay ready.

Updated 14 Nov 2026

01 / PROOF

The measured part

31
production deploys on Team RVM since 16 Oct 2026, all green
Source
Probes
health, data, TLS and a real API response after cutover, then rollback on failure
Source
3,640
Scheduler backend tests gating the deploy
Source

02 / WHAT I DO

In practice

Put the test suite in front of the deploy, not after it.
Probe the live system after cutover: health, data, TLS and a real API response.
Keep the previous images tagged and restore them automatically when a probe fails.
Make the rollback path testable, not theoretical.

03 / WHERE I'VE DONE IT

The systems behind it

Team RVMTests and a health wait gate every merge; 31 production deploys since 16 Oct 2026, all green.StrikeHikeTests, build, probes and automatic restore of the previous images.RVM CAD SchedulerBackend deploy lints migrations and runs the whole test suite first.

04 / STACK

What it runs on

GitHub Actions
Test, build and deploy pipeline
See the system
Docker
Compose stack built and cut over on deploy
See the system
Caddy
TLS, routing and reverse proxy
See the system
Nginx
Reverse proxy; config tested before reload
See the system
PM2
Process supervision
See the system
Prometheus (in my roles)In my roles
Prometheus in my roles (self-reported)
Grafana (in my roles)In my roles
Grafana in my roles (self-reported)

PR → production

PR
Tests
Build
Deploy
Probes
Rollback on failure

The release path on StrikeHike: a failed probe restores the previous images and fails the job.

05 / LIMITS

What this does not claim

No on-call rotation, SLO ownership or multi-region failover is claimed.

06 / RELATED CASE STUDIES

All work

ProductionStrikeHike

Live at strikehike.com; this engine replaced the first version on 25 Oct 2026. No accuracy or usage figures are claimed.

StrikeHike: a macro and planetary research terminal, built and run solo

A live research site that sets the US macro record beside the planetary sky, with tested claims, a weekly roundup, a public ledger, upcoming dates, and recession and decline histories. Next.js and FastAPI. Research, not advice.

131 of 131 commits on its main branch are mine

07 / QUESTIONS

Short answers

What happens when a StrikeHike deploy fails?

The workflow tags the running backend and frontend images before the build. If a probe fails after cutover, it restores those images, probes again, and fails the job because the new release did not pass.

Can the rollback itself be tested?

Yes. The deploy workflow has a manual rollback rehearsal input: it deploys, forces the probes to fail, and passes only if the previous images are restored and pass every probe.

Need this on a real system?

Hiring for a role, or scoping a short engagement? The Hire page has the role families and contact; Consulting has the scoped offers.

For hiring managers Consulting