Public teardown
I ran the diagnostic on [open-source project]’s public pipeline. Median time: [X] minutes.
Three findings:
- [finding one]
- [finding two]
- [finding three]
The fix is a merged pull request: [link]
CI/CD pipeline speed · Fixed price · Guaranteed
Your CI was sized for humans.
AI coding tools doubled your pull request volume. Your pipeline didn’t get faster. I halve median pipeline time for engineering teams of 100 to 500 developers in three weeks, or the second half of the fee is refunded.
Read-only. Two days. You get minutes lost per developer per day, in euros, before any money changes hands.
5+ years enterprise DevOps · GitHub Actions, GitLab CI, Azure DevOps · Based in Romania, working UK and EU hours
Why now
Nobody inside your company owns pipeline speed. It’s everyone’s annoyance and no one’s job. For three weeks, it’s mine.
What is your pipeline costing you?
Adjust the numbers. The defaults are typical for a 150-developer team.
How this is calculated
hours waiting per year = developers × runs × minutes × 0.33 ÷ 60 × 220
€ per year is hours × loaded hourly cost. Recovered is half of that. Payback is €12,000 divided by the daily recovery over 220 working days.
Assumes developers actively wait on a third of runs. Context-switching cost is excluded. Conservative on purpose.
- € per month
- €169,400
- € per year
- €2,032,800
- Hours waiting per year
- 25,410
- € recovered per year after halving
- €1,016,400
Your engagement pays for itself in 3 working days.
How it works
-
Week 1
Baseline and quick wins.
Access, measurement, then runner sizing, caching, queue elimination, obvious parallelization. Usually 30 to 40% off by Friday.
-
Week 2
The hard part.
Test selection and parallelization, flaky-test quarantine, merge queue, iteration caps for AI agents, security gates that don’t add a minute.
-
Week 3
Nothing changes.
I measure, hand over the dashboard, runbook and policies, and deliver the before/after number.
One CI system (GitHub Actions, GitLab CI or Azure DevOps), one main pipeline, up to 150 developers. Larger scope is a second tier, priced after the diagnostic. No surprises, because I look before I price.
Proof, not promises.
Sample report
Sample report, page one. Add public/media/sample-report-page-1.png and public/sample-report.pdf
This is what you get two days after access.
Before
42 min
After
19 min
Use the runners. I do. They’re the first 40%.
| What you get | Drop-in runners (Blacksmith, Depot) | DevOps agency | Me |
|---|---|---|---|
| Runner speed and cache | Yes | Yes | Yes, and I’ll recommend one |
| Test suite, flakes, merge queue | No | Sometimes | Yes |
| Measured before/after number | No | Rarely | Always, from your CI’s data |
| Guarantee | No | No | Second half of the fee refunded |
| Who does the work | Your team | Whoever is free that month | The person on this page |
| Time to result | An afternoon, for the first 40% | 6 to 12 weeks | 3 weeks |
| Price | Monthly minutes | A quote | Fixed, €12,000 |
Price and guarantee
€12,000, fixed.
What you pay
Half on signing. Half when the measured result is delivered.
Founding-client rate available for the first three clients, in exchange for a case study and a reference. Ask.
How the result is measured
- Metric: median pipeline duration on the main branch, queue time included.
- Source: your CI system’s own run data, exported. Not my dashboard.
- Windows: the five working days before week 1, and the five working days of week 3.
- Target: after is 50% of before, or less.
- Miss: the second half is refunded within five business days. You keep every change that shipped.
What I touch, and when I let go.
- Diagnostic: read-only access to CI configuration and run history. Nothing executes.
- Engagement: a named, scoped account on your CI and runner infrastructure. No production credentials, ever.
- Every change ships as a pull request your team reviews and merges. Nothing lands without a reviewer.
- Access is revoked at handover, and I confirm the revocation in writing.
- Diagnostic exports are deleted 14 days after the report. I keep only the report.
- I’ll sign your NDA.
For and not for
This works when
- You have 100 to 500 developers.
- You’re on GitHub Actions, GitLab CI or Azure DevOps.
- Your pipeline takes over 15 minutes.
- Your team already uses AI coding tools.
- Someone can grant access within a week.
This isn’t for you if
- You have under 30 developers. Switch runners; you’re done.
- You’re on Jenkins and need a migration first. That’s a different project.
- Your procurement cycle is longer than the engagement.
- You want a report instead of changes.
Photo. Add public/media/andrei.jpg
Andrei Badescu
[X] years running CI/CD, Kubernetes and cloud infrastructure for [industry 1] and [industry 2] companies at enterprise scale. Pipeline speed was always the thing everyone complained about and nobody owned. I decided to own it.
I work alone, so you get the person who does the work, not an account manager. Based in Romania, working UK and EU hours.
Video, 60 to 90 seconds. Add public/media/intro.mp4 and public/media/intro-poster.jpg
FAQ
Can’t we just do this with Claude Code or Copilot?
For the YAML, yes. The result comes from the test suite, the flakes, the runner fleet and the policies, which need access, judgment and someone accountable for the number. I use AI tools myself; that’s why it takes three weeks and not eight.
We already switched to Blacksmith, Depot or self-hosted runners.
Good. You’re the ideal client. The runner was the easy 40%; I do the rest.
What access do you need?
Read-only for the diagnostic. For the engagement, a scoped account on your CI system and runner infrastructure, revoked at handover. Every change goes through a pull request your team reviews.
What if you don’t hit the number?
The second half of the fee is refunded. You keep everything that shipped.
Do you work with Jenkins?
Not as a speed engagement. Jenkins estates usually need a migration first, which is a different project with a different price.
Our pipeline is already under 10 minutes.
Then you don’t need me, and I’ll say so after the diagnostic.
How soon can you start?
Diagnostic within two days of access. Next engagement starts [date]. I run two at a time.
Two days to a number. Three weeks to half.
Next engagement starts [date]. I run two at a time. I reply within one business day.