Fractional CTO · Freelance ML & AI engineer
Datooine is Thibault Djaballah, a machine learning engineer who built models at Criteo and Renault. Two ways to work together: a fractional CTO to set your technical direction without a full-time hire, or a senior ML & AI engineer to take your projects all the way to production.
Clients, missions and past teams
Depending on whether you need someone to decide, or someone to build. Often a bit of both.
For founders and executives who need a senior technical lead, without hiring a full-time CTO.
A few days a month, over time.
For teams with an AI or data project to deliver, who need a senior engineer who writes the code.
Daily rate or fixed scope, remote or on site.
A few situations where Datooine often comes in.
You have a product to build and nobody in-house to challenge your agency’s or freelancers’ choices, or to tell you what a feature should cost.
Developers ship, but without a clear architecture, tests or automated deployments. Every release is a gamble.
The prototype impressed everyone in the meeting. Six months later it still isn’t connected to your data or used by your teams.
Investors will ask hard questions about the tech. Better to get a clear-eyed assessment before they run their own.
One point of contact, who has spent as much time writing code as making technical decisions.
Recommendations are tested in code before they end up on a slide. When a prototype or a critical component has to ship fast, it gets written directly.
Ad bidding models at massive scale at Criteo, predictive maintenance at Renault, LLM agents on AWS today.
An LLM where it adds something, a simple rule or a SQL query where that’s enough. And you’ll hear it when AI isn’t the right answer.
Code in your repositories, infrastructure as code, up-to-date documentation: your team can carry on without Datooine.
Thibault, from the first call to delivery. No sales rep, no junior staffed behind the scenes.
Used to international teams, remote or on site.
Recent project
recto is an open-source flashcard app for the terminal, in the spirit of Anki. When you import a deck, a model writes variants of every card. Each review picks one at random: you type your answer, the model grades it and explains what was missing, then the FSRS algorithm decides when the card comes back.
$ uv tool install recto$ export OPENAI_API_KEY=...$ recto # wines.toml: a deck is just questions and answers[[cards]]id = "chablis-grape"question = "Which grape dominates the white wines of Chablis?"answer = "Chardonnay"A senior technical leader who works with you a few days a month. They own technical decisions, the team and the roadmap the way an in-house CTO would, for a fraction of the cost of a full-time hire. It’s often the right setup before your team is big enough to justify a full-time CTO, or while you’re recruiting one.
It depends on the phase. More at the start, when there’s an assessment to run, people to hire and practices to put in place. Less once the team is running. The pace is agreed together and reviewed regularly.
Yes, when it helps: a prototype to settle an architecture decision, a critical component, setting up CI. But as a CTO the goal is to make your team autonomous, not to become indispensable.
Agents connected to your tools and data, search and information extraction across your documents, triage of incoming requests, scoring, forecasting. Always with measurable evaluation, guardrails, and human review when the decision matters.
Models and data can stay in your own cloud, in a European region, or on self-hosted open models when confidentiality requires it. Nothing goes to a third-party provider without your agreement.
Most of the time, yes. Datooine is based in Montpellier, France, and travels for the moments that matter: kickoffs, workshops, hiring. Work happens in English or French.
Through Datooine Technologies, a French company. Daily rate, or a fixed price for a defined deliverable, with a quote before anything starts.
You’ll hear it at the end of the first call. It happens, and it’s good news for everyone: you leave with an outside opinion, and nobody has wasted any time.
30 minutes to talk it through. At the very least, you leave with an outside opinion.