Experience

Criteo

Criteo is a global technology company providing innovative advertising and marketing solutions.

▹ Senior Machine Learning Engineer

Sep 2021 - Present

Paris Area, France

Enriched User Timelines in a Data-Intensive Environment

• Enriched billions of user timelines with additional product universal category data, enabling the creation of more sophisticated and effective machine learning models.
• Optimized a data-intensive Spark job, transforming it into an incremental process that halved infrastructure costs.

Lookalike Audience Expansion Engine

• Built a lookalike audience expansion engine utilizing Faiss (a vector search library), leveraging existing user embeddings to meet client-specified target audience sizes.
• Integrated the engine with two of Criteo’s marketing solutions, achieving significant client adoption and seamless scalability with minimal maintenance over 1.5 years.

Embedding Quality Evaluation and Improvement

• Developed and enhanced homogeneity metrics for embedding quality, impacting hundreds of millions of products and billions of users.

360Learning

360Learning is a collaborative learning platform with interactive tools and analytics for corporate training.

▹ Machine Learning Engineer

Feb 2021 - Aug 2021, 7 mos

Full remote, France

• Built a multi-tenant learning module recommendation system as the company’s inaugural and sole ML Engineer.

Bedrock Streaming

Bedrock delivers cutting-edge video streaming platforms to broadcasters and media companies in Europe for more than 35 million users.

▹ Tech Lead, Personalization

Sep 2020 - Feb 2021, 6 mos

Paris Area, France

• Built Recommendation team of 5 people including engineers, product owner and UX designer, hired and mentored data scientists and data engineers.
• Coordinated Data Science and Backend teams to build end-to-end recommendation systems.
• Led system architecture design for ML applications.
• Developed deep neural networks using TensorFlow for batch and streaming user-item recommendation models for 35 million users.
• Built pipelines including databases, cache and API for recommendation model serving and evaluation.

▹ Senior Machine Learning Engineer

Nov 2018 - Sep 2020, 1 yr 11 mos

Paris Area, France

• Built a cross-platform (desktop, mobile and TV) lookalike audience extension system for 35 million users using classification models in Spark, which allows creating user segments up to 6 times larger than the base audience.
• Solved classification problems from positive and unlabeled data.
• Reorganized teams to improve cohesion between data scientists and data engineers, and mentored entry-level data scientists.

Etsy

Etsy is a global marketplace connecting 2 Million active sellers and 35 Million active buyers. Etsy contributes to open source software at the same time.

▹ Machine Learning Engineer

Apr 2014 – Sep 2017, 3 yrs 6 mos

Paris Area, France

Search Ranking

• Created the first Machine Learning powered application on alittlemarket.com (now etsy.com) for millions of unique items.
• Designed and developed a user behavioral data collecting system with Node.js and Elasticsearch.
• Developed a unique item purchase predictive model in Python(scikit-learn) and an Algolia reranking process.
• Deployed the predictive model in production as API using Flask.
• Designed and monitored Key Performance Indicators in Google Analytics and internal tools.
• Increased revenue by 15% (A/B test).

Recommender System

• Implemented a content-based recommender system with Spark Scala for millions of unique items.
• Parallelized batch jobs by using MapReduce programming model.
• Reduced job duration by optimizing algorithm’s time complexity.

Autocomplete Suggestions

• Built a pipeline in Python for extraction and spelling correction of e-commerce expressions in French.
• Built a parallelization tool to reduce processing time with Python multiprocessing.

Tech Stack

• Programming Languages and Frameworks: Python (Scikit-learn, numpy, scipy, pandas, Spark, Faiss, PyTorch, TensorFlow, Keras, matplotlib, Flask, Django); Scala (Spark, zeppelin)
• AWS (Sagemaker, S3); Databricks; MySQL; Hive; Cassandra; Redis; Docker; Airflow; Jenkins; Ansible; Grafana; Kibana; Algolia

Projects

Baby health (2017): Created a web application that calculates a 0-5 year old baby’s Body Mass Index and ranks their height and weight with babies around the world using World Health Organization open data.

Home intelligence (2016): Built an automatic lighting control system using Internet of Things.

Additional Experience and Awards

Top 6% (bronze medal), Kaggle competition, “M5 Forecasting - Accuracy”, 2020

• Created a time series forecasting algorithm that accurately estimate the unit sales of Walmart retail goods.

Finalist, Meilleur Dev de France

• Participated in Meilleur Dev de France 2018, an algorithm hackathon, being one of the 140 finalists out of 2000 developers.

Top 6% (bronze medal), Kaggle competition, “Predicting Red Hat Business Value”, 2016

• Created a classification algorithm that accurately identifies which customers have the most potential business value for Red Hat based on their characteristics and activities.

Education

Sorbonne University, Pierre and Marie Curie campus (Jussieu campus)

Master’s degree, Artificial Intelligence and Decision, 2014