1 of 27

Ship Better Code, Faster:

Optimize Developer Workflows with CI/CD and AI

Session 1

2 of 27

15 min: Introductions & my path into platform engineering
5 min: How the workshops work + expectations
15 min: The software delivery lifecycle and why companies hire for it
10 min: Clone the OrbitTasks repo onto your machine
15 min: Measure the current pipeline, your first deliverable

Agenda

3 of 27

Hi, I'm Riain.�Your guide for the next 8 weeks.

What I do

Platform Engineer at PermitFlow.

Background: Lead Platform Engineer @ Personio, SWE + Architect @ Aviva Ireland, �Computer Science @ Dublin City University

I've worked across frontend, backend, platform tooling, and AI-assisted engineering. Spent the most of my time caring about how teams ship, not just what they ship.

Communication + some more about me

You can ask me about: why platform, what the difference between "product" and "platform" really feels like, or how to talk about your work in an interview, career growth etc. - or anything else!

You can reach me via email.

I also love skiing and theatre, where I work internationally on site-specific work.

4 of 27

Your turn, who are you?

Round-robin, 60 seconds each.

Pick any two of the prompts below and share with the group. We'll go in the order Zoom shows you.

Your name, school, and what you're studying

Why you applied to this project specifically

A piece of software you wish was less frustrating

One thing you want to walk out of these 8 weeks with

5 of 27

How these�eight weeks work

6 of 27

From measuring the problem to proposing the fix.

Every week you deliver one piece of a single, growing engineering proposal. By week 8 you've made the case end-to-end. Each week builds on the last. Here's the shape of the arc.

Weeks 1–2

Measure the current pipeline and find the bottlenecks.

Weeks 3–4

Build a real CI/CD pipeline and improve developer experience.

Weeks 5–6

Use AI coding tools well, then prove the improvements with metrics.

Weeks 7–8

Set engineering standards and present the whole thing to stakeholders.

7 of 27

Cameras on, ask anything

These workshops are interactive. I want to see your faces and I want you to see each other. Use chat or unmute. If you're stuck on setup, you're probably not alone, so speak up early.

We do the work together in session

The core work happens live, together, in the hour. You won't have to take it home. Outside of session, expect up to 2 hours a week of mostly optional polish or reading. Workshop time is where the work gets done, but an hour isn’t a whole lot!

How we work�together for 8 weeks.

8 of 27

What is the deliverable?

01. The Repo

Over the next eight weeks we will remediate many issues in a sample repo, of which you’ll have your own copy to work on. By the end of our time together, you’ll (hopefully) have a much more performant repo.

02. The Presentation

For the final workshop, you’ll be asked to put together a presentation for stakeholders (me) where you’ll utilise all of your learnings and findings to illustrate the impact of your work.

9 of 27

Coursework + the app’s site

01. Coursework Plans

Access all plans and detailed information at:

https://tbf-coursework.riaincondon.com

* This may be most helpful post completion of this course.

02. Live Site & Code

Explore the live site and app codebase at:

https://tbf.riaincondon.com

https://github.com/riain0/tbf-orbittasks

* Login information can be found in the README.md

10 of 27

The software�delivery lifecycle

11 of 27

What is the Software Delivery Lifecycle?

One sentence definition

The full path a single line of code takes from the moment an engineer writes it on their laptop to the moment a customer feels the result.

A bottleneck anywhere along this path slows down the whole team, not just the engineer who wrote the code.

12 of 27

Code → Customer: The Lifecycle Stages

Every stage can become a bottleneck, and today we measure all of them.

code

Code + Review

An engineer writes a change and a peer signs off on it.

build

Build

The codebase compiles into runnable artifacts.

fact_check

Test

Automated checks run against the build to ensure quality.

rocket_launch

Deploy + Monitor

Ship to customers and watch how it behaves in production.

13 of 27

Where engineering time goes

What the data shows

Stripe's Developer Coefficient (2018): engineers spend ~42% of their week on maintenance, debugging, and bad code.

Atlassian's State of Developer Experience (2025): developers spend only ~16% of their time actually writing code. Half lose 10+ hours a week to friction; finding information, context switching, waiting on tools.

DORA's State of AI-assisted Software Development (2025): AI boosts throughput but hurts stability where foundations are weak. It amplifies your system of work, it doesn't repair it.

AI made writing code faster. It made everything around the code the bottleneck.

That's the problem this series is about.

14 of 27

settings

DevOps Engineer

Owns pipelines, deployment infra, and monitoring.

layers

Platform Engineer

Builds the internal product other engineers use to ship.

speed

Site Reliability Engineer (SRE)

Keeps the system up, defines SLOs, runs incident response.

Who gets paid to do this work?

The skills you'll build over these eight weeks map directly to three well-paid engineering roles. The work overlaps heavily, but the core focus differs.

All three start with the same question: where is engineering time being lost?

15 of 27

A look behind the curtain.

Four real engineering organizations, ordered from extreme to typical.

Netflix

4,000+ deploys per day. Every engineer can ship to prod safely.

Stripe

~10 minutes from commit to production. Internal tools are a first-class product.

Shopify

Their internal "Spin" platform boots a full dev env in under a minute.

OrbitTasks

A ~13-minute CI, flaky tests, toil

16 of 27

Meet�OrbitTasks

17 of 27

OrbitTasks: the situation.

28-person Series A startup. Project management SaaS.

~13 minutes

CI pipeline duration on every PR. ~95% is one stage. You feel every run.

~30% Failure Rate

Share of test runs that fail for non-deterministic reasons.

2 Days Onboarding

The time it currently takes to get a new engineer fully set up.

The Mission

You: the platform engineer hired to fix all of these critical issues.

18 of 27

Tool installation

Follow the instructions below for your operating system to get your environment ready.

macOS & Linux

Node & npm: Use nvm (Node Version Manager). Run the install script from nvm.sh then nvm install 24.
Git: macOS usually has it. If not, run brew install git. Linux: sudo apt install git-all.

Windows

Node & npm: Download the v24.x.x LTS installer (.msi) directly from nodejs.org.
Git: Download and run the "Git for Windows" installer from git-scm.com. Use the default settings.
Terminal: Use PowerShell or Git Bash for the best experience.

19 of 27

Step 0: Verify your environment.

Run the four commands below to ensure your local machine is ready for the project.

Check these versions in your terminal:

node --version: v24.x.x or newer.
npm --version: v10.x.x or newer.
git --version: any 2.x release is fine.
macOS/Linux:

ls: should show apps/, scripts/, package.json at project root.

Windows:

dir: should show apps/, scripts/, package.json at project root.

If any command produces errors, say so now. Fixing setup live is part of the work.

20 of 27

Run these commands

git clone https://github.com/riain0/tbf-orbittasks.git�cd tbf-orbittasks

then

npm install

If something breaks

Drop the error message in chat. We will troubleshoot live; this is part of the work, not an interruption.

If your machine is too constrained, pair with a neighbor and we'll get you set up afterward.

Step 1: clone OrbitTasks locally.

21 of 27

Establish�the baseline

22 of 27

Three numbers that matter today.

Numbers without observations are noise. Observations without numbers are opinions. We want both. Capture all three categories below as you run the pipeline.

01. Wall-clock time

How long from "I push" to "this stage finishes."

02. Pass / fail

Did the stage complete cleanly? What was the visible error?

03. Observations

Anything that felt slow, redundant, or confusing. Gut reactions count.

We record all of this in your course logbook as we go. We reuse this data in Session 2.

23 of 27

Eight stages, one row each

Or: run npm run ci to do all eight in one go, we’ll do this in a second

01. install

npm install (~7–90s)

02. lint

npm run lint (~2s)

03. typecheck

npm run typecheck (~2–3s)

04. test:api

npm run test --workspace=apps/api (~12 min)

05. test:web

npm run test --workspace=apps/web (~2–10s)

06. build:api / web

npm run build --workspace=apps/... (~1s each)

07. deploy

npm run deploy (~18s)

08. Logbook & Automation

Wrapper script writes timing to baseline.log automatically.

24 of 27

Run it. Watch it. Write it down.

We run npm run ci at the repo root. It takes ~13 mins (mostly test:api). We capture observations in your logbook as it runs.

01. Open

Open your terminal at the project root.

02. Run

Run npm run ci. It writes to baseline.log.

03. Capture

Fill the W1 logbook section. Note what surprises you.

04. Log

Fails are data. Write down what failed and how.

25 of 27

Duration of each stage. Total wall-clock time end to end. Pass or fail per stage. How long `npm install` spent fetching versus actually installing.

Which stage felt slowest? Did any test fail, and was it repeatable? Anything that confused you or looked redundant? Gut reactions go here.

Numbers

Observations

What goes in�your logbook.

26 of 27

Profiling and Root�Cause Diagnosis.

NEXT SESSION PREVIEW

We'll pick up the data you just logged and tear it apart together. Three things we'll cover next time:

analytics

Profiling techniques: finding the actual hot path in a pipeline, not just the obvious one.

account_tree

5 Whys & Fishbone: industry frameworks for getting past symptoms to root cause.

low_priority

Prioritization: which bottlenecks are worth fixing first, by impact and effort.

menu_book

Keep your logbook; we reuse today's baseline numbers next week.

27 of 27

Questions?

on the workshop, my experiences, anything i can help with.