CI/CD, Test Automation & Platform Consultant

Your CI was built for 10 engineers. Now you have 50, and their agents.

You hired more engineers and bought AI tools. Delivery did not speed up. I measure what slows you down and what it costs, then build the fix with your team: faster pipelines, safer releases, and AI agents where they help.

Théo Penavaire in a navy polo, holding a ceramic cup

Three steps. You can stop after the first.

For engineering teams of 20 people or more whose CI and tests slow them down. Each step has a fixed scope and a fixed length.

  1. Delivery Pipeline Audit

    2–3 weeksStart here

    For
    Teams that feel the slowdown but cannot say yet where it comes from.
    You get
    A CI audit of your pipeline and test suite: measured wait times, flaky tests and coverage gaps, the cost of each bottleneck, and a ranked list of fixes.
    Ends with
    A written report and a review session with your tech leads.
    Start with an audit

    How do you price?

  2. Implementation

    1–3 months

    For
    Teams that know what to fix and have no time to do it.
    You get
    The top fixes, built with your engineers: faster feedback, fewer flaky tests, a main branch that stays green, and AI agents in the pipeline where they save time.
    Ends with
    Before-and-after measurements and a handover to your team.
    Talk about implementation
  3. Embedded

    6–12 months

    For
    Teams that grow fast and need someone to own delivery for a while.
    You get
    A senior platform engineer inside your team who runs the delivery roadmap and the developer productivity work with you.
    Ends with
    A team that runs and improves its pipeline without me.
    Talk about an embedded role

Together, we analyse your bottlenecks

I use the same five steps on every pipeline, in this order. Most teams start at step five.

  1. Question each requirement

    We start from first principles. What does your team need to ship safely? Every check in your pipeline had a reason once. We keep the ones whose reason still holds, and give each one an owner.

  2. Delete what has no reason

    Jobs nobody reads, tests that never fail and steps kept "just in case" go first.

  3. Simplify what stays

    Fewer stages, clear owners, one way to do each thing.

  4. Make it faster

    Then we cut the wait with caching, test selection, parallel runs and merge queues.

  5. Automate last

    Automation, agents included, comes last. It goes on the process that survived the first four steps.

What agents can't do for you

I use AI agents in my own work every day, and I add them to delivery pipelines where they save time. People still add the most value in three places.

Context
Why your system looks the way it does, and what the business needs from it next quarter.
Verification
Proof that a change is right, beyond a green check.
Taste
The choice of what to build, and what to leave out.

Missions

Where this method comes from, and what the people I worked with say about it.

  • Six years on the CI and test systems behind the Unity engine. I moved the Graphics SRP codebase (500 developers, 30,000 commits) into the engine repository with no data lost and no development stop, and test coverage grew 50 times.

    • 500 developers on the migrated codebase
    • 30,000 commits migrated
    • ×50 test coverage
    • Migration
    • Test automation
    • Test selection
    • 20+ platforms

Recommendations

  • He has a rare ability to combine strategic thinking with consistent, high-volume delivery — he doesn't just identify what needs to be done, he follows through and gets it done, thoroughly and reliably.

    Jan Borup JakobsenSenior Manager, Automation and Tooling at Unity.
  • He's had large impact on Unity's graphics related CI pipelines and tooling, enabling this important area of the Unity Engine to cover all the relevant dimensions of build configurations at scale. […] He's also been an early adopter for AI agentic development, improving his productivity even further.

    Henrik KjellanderStaff Software Engineer, Release & Engineering Services at Unity
  • He has a deep understanding of CI/CD and build systems, but what really sets him apart is his ability to make complex processes easy to understand.

    Alex Vázquez MartínezSenior Software Engineer at Unity

All missions

Tools & certifications

Daily work
  • Git
  • GitHub
  • TypeScript / Node.js
  • Playwright
  • Vitest
  • AWS
  • Claude / Codex
  • Python
  • C# / .NET
  • Bash
  • Docker
Also worked with
  • GitLab
  • Jenkins
  • Selenium
  • GCP
  • Azure
  • Langchain
  • LiteLLM
  • Unity
  • C++
  • Grafana
  • BigQuery
  • SQL
Certifications

Different CI system or tech stack? The bottlenecks are usually the same. Ask me.

Writing

What I learn about building and shipping software, written down as I go.

  1. 13 min read

    How many changes can you land at once?

    Merge queues automate rebasing and testing PRs at scale, with batching, speculative checks, and bisection tradeoffs explained through failure-rate math.

  2. 7 min read

    The tests you didn't need to run

    Fix slow CI with test selection: test impact analysis, only-changed flags and Meta's Predictive Test Selection skip the tests a change can't break.

All posts

About

I'm Théo, a freelance CI/CD and test automation consultant. For six years I kept CI and test automation working for Unity's Engine organisation, on more than 20 platforms, from mobile to XR and consoles.

I lived in Paris for 30 years, and now operate from Norway. I work remotely with teams across Europe, in English or French.

English · French · Spanish (intermediate, B1) · Norwegian (elementary, A2)

More about me

A 30-minute call can change your delivery roadmap.

No commitment and free. Tell me where delivery hurts, and I'll tell you if I can help.

Book a free 30-min call