> Markdown version of [/jobs/ext/1341357-data-scientist](https://www.wearedevelopers.com/jobs/ext/1341357-data-scientist). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Data Scientist - **Company:** Vibe - **Location:** Paris, France - **Experience:** Expert - **Contract:** Permanent contract - **Skills:** Training Data, Airflow, Information Leak Prevention, Software Debugging, Python (Programming Language), Tensorflow, Pytorch, Deep Learning, Kubernetes, ONNX (Open Neural Network Exchange) Format - **Published:** July 19, 2026 - **Apply:** https://fr.indeed.com/viewjob?jk=17fa14408151dff8 ## About the Role * Real hands-on deep learning experience in a professional environment * Strong Python; PyTorch preferred (TensorFlow or JAX is fine) * Comfort diagnosing why a model works or fails: data leakage, bias, calibration, distribution shift, optimization issues * Ability to build algorithms from scratch and reason about what's under the hood * A track record of framing your work in business outcomes: you can point to the KPI you moved and by how much Nice-to-Haves * Based in Paris or open to relocating (hybrid, 2-3 days in office); relocation support available * Worked with massive-scale datasets: billions of impressions, events, user data * Deep learning on tabular data and sparse user representations * Ad tech experience, especially CTR/CVR prediction or recommendation models in a DSP * Familiarity with identity graphs or cross-device attribution * Adjacent domains with large-scale decision-making on rich tabular data: retail, travel and dynamic pricing (booking, airlines), systematic trading * Production ML tooling: ONNX export, orchestration frameworks (Dagster), inference serving (Triton) ## Description You'll join the Performance team, the engineers and scientists who own advertiser outcomes at Vibe: prediction models, bidding, and the delivery stack that drives both. This role exists because our modeling roadmap is bigger than our current bandwidth. You will help to build the core prediction architecture that powers every bid, and the scale and diversity of data available to us is about to grow significantly. We need someone who can turn that into measurable advertiser performance, not just publishable ideas. Reasons to join: * CTV performance advertising is still being invented. Cross-device attribution, identity graphs linking TVs to households - the modeling problems are nothing like web display or search. There's no playbook to copy; you'd be writing it * You'll work on data few others have. We're building toward crossing CTV signal with real-world purchase and behavioral data: joining what people watch with what they actually buy. Measurement, incrementality, training data: all of it changes * The timing is right. Vibe is one of the fastest-growing CTV companies out there. Early enough that the model architecture is yours to shape. Late enough that your work ships to real advertisers and shows up in the numbers within weeks What You'll Do: * Own the prediction stack * Improve our unified multi-task model: one architecture predicting visits, purchases, and other outcomes simultaneously, with heads learning from each other through shared representations to handle sparse conversion funnels and enable more precise bidding decisions * Take ideas from exploration to production yourself: dataset design, architecture, training, debugging (gradients, convergence, etc.), deployment, and monitoring for drift and regression. You will work closely with Performance engineers building the serving and bidding infrastructure * Design new features from raw signal, including household-level features from our identity graph * Prepare the modelling stack to absorb new, richer data sources * Optimize for business impact: measurable uplift for advertisers over offline metrics that never ship * Iterate in production, not in your head: ship, measure, adjust ## Related Videos - [Data Science in Retail](https://www.wearedevelopers.com/videos/586-data-science-in-retail) - [Machine learning in the browser with TensorFlowjs](https://www.wearedevelopers.com/videos/155-machine-learning-in-the-browser-with-tensorflowjs) - [From Messy Queries to Scalable Systems - How Data Engineering actually works](https://www.wearedevelopers.com/videos/100203-from-messy-queries-to-scalable-systems-how-data-engineering-actually-works) - [Photonic Computing: Programming a New Class of AI Accelerators (incl. Live Coding)](https://www.wearedevelopers.com/videos/100196-photonic-computing-programming-a-new-class-of-ai-accelerators-incl-live-coding) - [How We Built a Machine Learning-Based Recommendation System (And Survived to Tell the Tale)](https://www.wearedevelopers.com/videos/752-how-we-built-a-machine-learning-based-recommendation-system-and-survived-to-tell-the-tale) - [Geometric deep learning for drug discovery](https://www.wearedevelopers.com/videos/264-geometric-deep-learning-for-drug-discovery) ## Related Articles - [Coffee with Developers - Maria Apazoglou - Making AI understandable for all in production](https://www.wearedevelopers.com/magazine/475-coffee-with-developers-maria-apazoglou-making-ai-understandable-for-all-in-production) - [Highest Paying Tech Companies for Developers](https://www.wearedevelopers.com/magazine/220-highest-paying-tech-companies-for-developers) - [Dev Digest 120 - Apple and peers](https://www.wearedevelopers.com/magazine/455-dev-digest-120-apple-and-peers) - [Navigating the AI Shift](https://www.wearedevelopers.com/magazine/629-navigating-the-ai-shift) - [MLops – Deploying, Maintaining And Evolving Machine Learning Models in Production](https://www.wearedevelopers.com/magazine/115-mlops-deploying-maintaining-and-evolving-machine-learning-models-in-production) - [Dev Digest 193: Vibe Coding Honeymoon, NaN and the End of Interviews](https://www.wearedevelopers.com/magazine/652-dev-digest-193-vibe-coding-honeymoon-nan-and-the-end-of-interviews)