Criteo Research is pioneering innovations in computational advertising. As the center of scientific excellence in the company, Criteo Research delivers both fundamental and applied scientific leadership through published research, product innovations and new technologies powering the company’s products.
The Criteo Research team operates within the spectrum of two main roles:
Applied Research: Criteo Research 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.
A sampling of the research topics we work on in the above roles:
- Click prediction: How do you accurately predict if the user will click on an ad in less than a millisecond? Thankfully, you have billions of data points to help you.
- Recommender systems: A standard SVD works well. But what happens when you have to choose the top products amongst hundreds of thousands for every user, 2 billion times per day, in less than 50ms?
- Auction theory: In a second-price auction, the theoretical optimal is to bid the expected value. But what happens when you run 15 billion auctions per day against the same competitors?
- Bandit Algorithms: It’s easy, UCB and Thomson sampling have low regret. But what happens when new products come and go and when each ad displayed changes the reward of each arm?
- Reinforcement learning: How to find the optimal bidding strategy across multiple auctions? Can this be cast as a reinforcement learning problem in very high dimensions with unpredictable rewards.
- Offline testing/Metrics: You can always compute the classification error on model 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?
- Scalable Optimization: Stochastic gradient descent is great when you have lots of data. But what do you do when all data are not equal and you must distribute the learning over several hundred nodes?