I have been building mobile apps since 2018. Forty-four projects later, for startups as well as Bayard or Mastercard, I founded Naelo: a house of my own, where I build my own apps and welcome other people's projects.
I set up a workshop there, where AI agents work in a chain under my control. They are fast. I make sure it is done well.
Make yourself at home.
Aurélien
Freelance CTO and founder of Naelo Studio
The workshop
Chatting with an AI is not working with an AI.
A well-kept house does not rest on one talent. It rests on gestures repeated in order by hands that know their job, and on someone who comes round to check. My workshop runs that way: the AI agents make the gestures, I keep the house.
Everyone has asked a chatbot a question. Agentic development is something else: a mission is handed to several agents, each with a precise role, working one after the other like the stations of a workshop. The work moves forward without me, never without checks.
- 1Brief
You describe the need. I turn it into a precise specification that you approve.
- 2Plan
I split the work into small independent tasks, each with its own pass criterion.
- 3Tests
A first agent writes the tests before a single line of code.
- 4Code
A second agent writes the code until every test passes.
- 5Review
A third agent, which saw nothing of the writing, reviews and flags what is off.
- 6Checks
Automatic verification at every step. Everything must be green to move on.
- 7Delivery
I read, I approve, I ship. A human always signs the release.
Chatting
- One question, one answer, then you copy it over by hand.
- Nobody checks what the AI wrote.
- Quality depends on the mood of the day.
- Everything stops when you close the window.
Orchestrating
- A whole mission, split into tasks, carried from start to finish.
- Each agent is reviewed by another one that never saw its work.
- The same checks on every pass, no exceptions.
- Repetitive tasks run on their own, overnight if needed.
The rules that are not up for discussion
- Tests first
- We write down what “it works” means before writing the code. No agent may weaken a test to make it pass.
- Fresh eyes
- The reviewer is never the author. The reviewer gets the result, not the excuses.
- Two rounds, no more
- If a problem survives two fixes, the chain stops and I take over.
- Nothing ships in red
- A failing check blocks everything. There is no “we will see later”.
What runs on its own
A chain builds things. Workflows take care of everything that comes back every week.
- Sorting the errors users run into and proposing a fix
- Preparing App Store listings and screenshots
- Watching an app's market and competitors
- Tracking ad campaigns and cutting what does not pay
The estate
The apps of the house, grown here from the first idea to the App Store.
- Reselling
ResellStats
What you really make on Vinted, Depop and eBay.
Stock, net margin after fees and shipping, and the amounts French resellers have to declare.
Visit the site
- Strength training
Rest
The rest timer that starts by itself.
Log your set and the rest starts with the right duration for the exercise. On the lock screen, no account.
Visit the site
- Documents
Paraph
Sign your PDFs on iPhone.
Signature, initials on every page, place and date. Documents never leave the phone.
Visit the site
- Game
Darts Counter
Darts scoring without the mental maths.
501, 301, Cricket, Shanghai. You throw, the app counts and calls the end of the game.
Visit the site
The guest table
I take on few projects at a time, so that each one gets real attention.
Launch your app
From the idea to the App Store and Google Play: mockups, development, payments, release. A first usable version within a few weeks.
Rescue an app that is stuck
It crashes, it lags or it fails store review? A quick audit, the essential fixes, and clean documented code at the end.
Take your team agentic
I set up a chain of AI agents built on your existing tests and checks, and train the team to run it. That is what I do at Bayard.
Fractional technical direction
Technical choices, pace, quality: I act as your CTO a few days a month.
The guest book
- 44
- projects delivered on Malt
- 5/5
- average over 32 reviews
- 2018
- first year as a freelancer
- Bayard, La Croix app
Three times more tickets shipped, 40% fewer errors, load time down from 3.2 to 1.8 seconds.
- Mint
Stability up from 40% to 97% in one week.
- The Citadel
Load time divided by three and an app approved on both stores.
They trusted me with their product
Mastercard, Bayard, Epsor, Skin Easy, Avis 2 Santé, Poolp, Epitech