Boris Prodhomme
Associate Director Data Science & Machine Learning
Salaire / Taux journalier
Paris, France
Freelance
Peut venir sur site
Expertises
Langues
À propos
Introduction
AI/ML leader with 10+ years delivering production-grade systems at scale. Currently driving $100M+ UA optimization at Homa—slashed LTV prediction error by 50%, automated 25% of ad spend with sustained ROAS ≈1.2. Expert in building ML platforms, leading cross-functional teams, and shipping measurable business impact through predictive modeling and MLOps. Let's turn your data into revenue.
Expérience de travail
Associate Director Data Science, Machine Learning & AI
Homa
nov. 2023 - Présent
3 ans
Paris, France
Led Data Science & ML to production at scale. Managed a team of 4 DS + 1 MLE, driving predictive modeling, automation, and ML platform engineering. Delivered measurable business impact across LTV prediction, UA automation, and ML infrastructure. ✨ Key Impact - Leadership → recruited & coached DS/MLE; built strong practices (testing, alerting, documentation). - LTV Prediction → daily retraining (forward-fitting + LightGBM); cut wMAPE from 30%→15% on $100M+ spend; predictions available by 08:00 CET; full adoption by UA & exec teams. - Autobidder → refactored dbt code into reliable, modular system; automated ~25% of ad spend across multiple networks & campaign types; sustained ROAS ≈ 1.2. - ML Platform → designed modern stack (ZenML, Databricks, SageMaker, MLflow); standardized feature/training/inference pipelines; reduced ML MVP time-to-market to ≤ 1 quarter.
Head of Data
Immortal Game
oct. 2022 - oct. 2023
1 an 1 mois
Paris, France
- Built a Data team composed of 2 analytics engineers and 1 data analyst - Set up the Data Platform for efficient Data Analysis and AI workflows. - Hands in the dirt deploying Anti-Cheat algorithms, crucial for a chess platform. Set up a scoring system for all Cheating behaviors, and a prototype for Engine Use Detection based on Tensorflow. - Tech : BigQuery, dbt, Python, Tensorflow, Segment, Amplitude, Tableau, Husprey
Equativ (formerly Smart)
mai 2013 - mai 2021
8 ans 1 mois
Paris, France
Staff Data Scientist
juil. 2020 - mai 2021
11 mois
Paris, France
In charge of : - Managing all Data Science interns - Transitioning the two bid throttling (classification) algorithms to a new standard pipeline using Google Cloud Platform, Tensorflow, Tensorflow Serving - Detecting new ML applications (e.g. click rate prediction, pacing algorithm) for the newly acquired LiquidM company as well as organizing knowledge transfer sessions between both DS teams - Tech stack : Google Cloud Platform (Dataflow, Google Cloud Storage, BigQuery, AI Platform), Tensorflow, Tensorflow Serving, Docker
Senior Data Scientist
juil. 2017 - juil. 2020
3 ans 1 mois
Paris, France
Managing cross-functional projects driven by the Data Science team, from research to production. More precisely I was in charge of : - Two Bid Throttling algorithms absolutely core to the company business : binary classification and regression problems both solved by custom probabilistic multi-way decision trees. - Maintenance and new usages for the Ad Inventory Forecast module as well as reducing technical debt by transitioning data querying to ClickHouse. - Designed and implemented the ML part of a Video Ad Revenue Optimization algorithm using a data pipeline built on Hadoop, XGBoost and ONNX runtime - Designed and implemented a multi-dimensional revenue anomaly detection system - Management of Data Science 6-months interns - Tech stack : Hadoop, Spark, Parquet, C#, Python, XGBoost, ONNX, Tableau, Git
Data Scientist
mai 2013 - juil. 2017
4 ans 3 mois
Paris, France
Smart AdServer is an AdTech platform delivering more than 200 billion ads per month, therefore facing intense big data challenges. I participated in building the new AI decisioning engine, and was involved in most mathematical projects : - High-dimensional ad inventory forecast and ad delivery control (time-series forecast, probabilistical modelling & highly-optimized C# code) - Multichannel revenue optimization - Automated webpages crawling and classification into big advertising categories (Sports, Food, Finance…) - Tech stack : C#, Python, SQL, Hadoop, Git
Research Assistant
California PATH
sept. 2011 - mai 2013
1 an 9 mois
2105 Bancroft Way Suite 300, Berkeley CA 94704
Designing and implementing algorithms for improving traffic estimation using GPS and Bluetooth data from cell phones. - Designed and built data fusion algorithms and parameter calibration framework involving fixed sensors and cell phones data incorporation. - Defined new metrics and developed Matlab data visualization tools assessing the quality of large-scale probe data sets.
Intern
Bouygues e-lab
avr. 2011 - juil. 2011
4 mois
40 rue de Washington - 75008 Paris FRANCE
Research intern (Bouygues is a French giant in construction and telecommunications) - Designed and implemented in C# four algorithms for optimizing forward planning of maintenance on road networks. - Performed a detailed quantitative analysis comparing memory usage, time efficiency and accuracy of these algorithms.
Formation
University of California, Berkeley
Master of Science (MSc), Industrial Engineering and Operations Research
2011 - 2012
2 ans
École Polytechnique
Diplôme d'ingénieur (master), Mathématiques Appliquées
2008 - 2012
5 ans
Lycée Clémenceau
Mathématiques et Physique
2006 - 2008
3 ans
Lycée Guy Moquet
Licences et certifications
Generative AI with Large Language Models
Délivré le : oct. 2023
AI for Medical Diagnosis
Délivré le : août 2023
Neural Networks and Deep Learning
Délivré le : juin 2022
Smart Contracts
Délivré le : janv. 2022
Blockchain Basics
Délivré le : oct. 2021
End-to-End Machine Learning with TensorFlow on Google Cloud
Délivré le : sept. 2020
Introduction to Deep Learning
Délivré le : mai 2019
Machine Learning
Délivré le : août 2014
Computing for Data Analysis
Délivré le : mars 2013
Data Analysis
Délivré le : févr. 2013