Research Scientist

Grenoble, France
Who we are
At Criteo, we connect 1.5 billion active shoppers with the things they need and love. Our technology takes an algorithmic approach to predict what user we show an ad to, when, and for what products. Our dataset is about 50 petabytes in Hadoop (more than 120 TB extra per day) and we take less than 10ms to respond to an ad request. This is truly big data and machine learning without the buzzwords. If scale and complexity excite you, join us. 

The Criteo AI Lab is pioneering innovations in computational advertising. As the center of scientific excellence in the company, we deliver both fundamental and applied scientific leadership through published research, product innovations and new technologies powering the company’s products.

The Criteo AI Lab operates within the spectrum of two main roles:

Applied Research: Our Scientists fully-leverage the advantage of working in a machine-learning driven organization by partnering closely with our product and engineering counter-parts to deliver cutting-edge solutions to the challenges in online advertising.

Academic Contributions: The Research Scientists at Criteo are encouraged and fully-supported to publish their works at international conferences, collaborate with academic partners, file for patents, release datasets and help establish the state-of-art in computational advertising.

Research Topics

    • Deep Learning: We ingest an enormous number of product catalogs (images, text descriptions, historical datas) that we need to normalize while maintaining interpretability. Throw in attribute identification & extraction at scale to build actionable product graphs.

    • Interpretable AI: Why do deep learning models work? Are there underpinnings in probabilistic theory, information theory or optimization that explains the generalizability of the class of deep learning models or explain their performance? We would love to figure out the answers to the above with you.

    • Design of Experiments/ Non-Parametric Statistics: You can always compute the classification error on models predicting the probability of a click. But is this really related to the online performance of a new model? What is the right offline metric that predicts online performance?

    • Optimization: Stochastic gradient methods are great when you have lots of data. However, what does one do when all data are not equal and you must distribute the learning over several hundred nodes?

    • Recommender Systems: Standard matrix factorization may work. But, can we use deep learning techniques to leverage both collaborative filtering signal and product meta-data? Also, could we better personalize our recommendations with user modelling via Recurrent Neural Networks and other architectures?

    • Reinforcement learning: How to find the optimal bidding strategy across multiple auctions? Can this be cast as a reinforcement learning problem with a very high dimensional state space and unpredictable rewards?

    • Auction theory: In a secondprice auction, the theoretically optimal strategy is to bid the expected value. But what happens when you run 20B auctions per day against the same competitors?

What you'll do

    • Understand and shape the product direction by contributing innovative ideas, specifically as a result of data mining and experimental data modeling
    • Influence the strategic vision for the research team and Criteo at large.
    • Identify new research opportunities at Criteo and lead the exploration of these ideas and pursue patents/publications where appropriate.
    • Interact with other teams to define interfaces and understand and resolve dependencies
    • Maintain world-class academic credentials through publications, presentations, external collaborations and service to the research community.
    • Develop high-performance algorithms, test and implement the algorithms in scalable and product-ready code.
    • Mentor team members, oversee the creation of technical documents and work towards establishing Criteo as a center of excellence in computational advertising.

Who you are

    • PhD in Machine Learning or a related field along with at least five years of experience.
    • Strong hands-on skills in sourcing, cleaning, manipulating and analyzing large volumes of data.
    • Strong implementation experience with languages, such as Python, Perl, Ruby, Java, C#, Scala etc.
    • Familiarity with Linux/Unix/Shell environments.
    • Knowledge of Hadoop programming environments (e.g. Pig, Hive).
    • Excellent track record in conducting and reporting results of original and collaborative research with publications.
    • Prior experience in optimization of online advertising is a plus.

At Criteo, we dare to be different. We believe that diversity fuels innovation and creates an energy that can be seen and felt all over Criteo. We champion different perspectives and are committed to creating a workplace where all Criteos are heard, feel a sense of belonging, and are treated with respect and dignity.

Criteo collects your personal data for the purposes of managing Criteo's recruitment related activities. Consequently, Criteo may use your personal data in relation to the evaluation and selection of applicants. Your information will be accessible to the different Criteo entities across the world. By clicking the "Apply" button you expressly give your consent.