Freelance consultant · Senior Data Scientist · Paris, France
Causal ML and experimentation for marketing teams.
I'm Naoufal, a data scientist based in Paris. Over the past seven years, PhD at École Polytechnique included, I've built machine learning that marketing and product teams actually use: A/B tests, uplift models and recommender systems, taken from the first notebook to production. I now do this as a freelance consultant.
Selected outcomes
- €1M+ Incremental revenue per month Air France-KLM · A/B-tested recommenders, +0.6% conversion and +0.5% revenue per visitor (relative) on hundreds of thousands of visitors a day · 2025–26
- Up to 30% Relative conversion uplift in digital campaigns Senzai · uplift-based targeting for a telecom operator and a bank, 3M customers and 50M interactions · 2023–25
- ICML 2023 Peer-reviewed research in causal machine learning Meta-learners for multi-valued treatments · plus a CSDA journal paper, a patent application and a PhD from École Polytechnique
Four projects, with the numbers that came out of them.
Personalising flight and fare offers with A/B-tested recommenders
Two recommender systems on the booking flow: which flights to show first, and which fare to suggest. I work on the metrics and segments, prepare the data, and run the offline and online evaluation of every A/B test before anything ships.
Uplift-based targeting for digital marketing campaigns
I led the machine learning of a re-engagement engine: the recommendation system that decides which customers to contact, when, and through which channel, built on causal inference rather than propensity, plus the pipelines, evaluation and CI/CD it runs on.
Causal inference and uncertainty quantification for energy production
Three years of research where statistical learning meets causal inference: calibrated prediction intervals for gas-well production, and treatment-effect estimation for geothermal wells. It led to a patent application and papers at ICML and in CSDA.
Evaluating frontier language models on mathematics, agents and data
Three projects for AI research labs: writing mathematics problems that frontier models get wrong and experts can solve, evaluating models and agentic frameworks under constraints, and a STEM review project producing data visualisations with the science checked behind each one.
What I can help with.
- 01
Experimentation
Setting up, or fixing, an A/B testing practice: which metric, how much traffic, how long, what could go wrong, and a clear answer at the end. I've done this on flows with tens of millions of visits a month, and with teams running their first test.
- 02
Uplift modelling and targeting
Finding the customers a campaign will actually change, rather than the ones who'd have converted anyway. Causal models, meta-learners, and the evaluation that shows they beat whatever targeting you use today.
- 03
Recommender systems
Ranking and offer recommendation on booking, checkout or CRM flows. Evaluated offline first, then against a control group, so nobody has to take the model's word for it.
- 04
Getting ML into production
Pipelines, CI/CD, model tracking and monitoring, with Airflow, MLflow, Docker and Kubernetes on AWS or GCP. I've led this end to end, and I'd rather work with your engineers than around them.
A few things I insist on.
We agree on the metric before anyone touches the data. Most of the value of an experiment is decided right there, and it's the step people most like to skip.
Ideas get tested offline first, then with a real control group. Models that look great on historical data fail in production more often than you'd think, so I don't ask anyone to trust a model, I ask them to trust the test.
Whatever we ship comes with pipelines and monitoring, so the effect we measured is still there three months later. I write things down, I'd rather show a plot than a slide, and I'll tell you when the honest answer is "inconclusive, let's run it longer".
Where I've worked.
- Nov 2025 – PresentSenior Data Scientist, Marketing Operations Research · Air France-KLM
A/B-tested recommender systems on the booking flow, in a team of five or six data scientists. I handle the metrics, the data and the offline and online evaluation.
- Aug 2025 – Jul 2026Mathematics and Statistics Expert · Mercor
Three evaluation projects on frontier language models, for AI research labs.
- Jan 2025 – Jul 2025CTO · Kemba AI
Built and launched the MVP of a multi-agent platform for financial services.
- Oct 2023 – Oct 2025Senior Data Scientist and Machine Learning Engineer, Pipelines and MLOps Lead · Senzai
Led the machine learning of an uplift-based re-engagement engine for a telecom operator and a bank, and ran it in production.
- May 2023 – Oct 2023Data Scientist, Geo and NLP team · namR
Feature engineering and pipeline optimisation on large geospatial datasets, on GCP.
- Oct 2019 – Dec 2022Research Engineer, AI and Data Science R&D (industrial PhD, CIFRE) · TotalEnergies One Tech
Industrial PhD on statistical learning and causal inference for energy production.
Research
- Comparison of meta-learners for estimating multi-valued treatment heterogeneous effects ICML 2023
- Statistical learning and causal inference for energy production PhD thesis, École Polytechnique, defended in November 2022
- Robust Prediction Interval estimation for Gaussian Processes by Cross-Validation method Computational Statistics and Data Analysis, 178:107597, 2023. doi 10.1016/j.csda.2022.107597
- From what to why: the rise of causal AI Elaia Partners · Quoted on causal targeting: “Instead of asking ‘Who is likely to churn?’, it asks: Who will churn if I don't act, but stay if I do?”
Recent talks
- Aug 2022Greek Stochastics μ' 2022: Causal Learning Talk · Corfu, Greece
- Jun 2022MASCOT-NUM 2022 annual meeting Talk · Clermont-Ferrand, France
- Jun 2022The Causal TAU seminar Invited talk · Inria · Gif-sur-Yvette, France
- May 2022Journée Causalité/XAI Talk · SINCLAIR, EDF Lab Paris-Saclay · Palaiseau, France
Got a measurement, targeting or personalisation problem?
I take on freelance and consulting work in marketing analytics, experimentation and causal ML, anything from a single A/B test to a full personalisation programme. I'm in Paris, France, happy to work remotely or on site, in English or French. The quickest way to find out whether I can help is a short call: send me an email and we'll set one up.