Data Scientist (Internship)
Teads: No.1 Video Advertising Marketplace Reinventing Digital Advertising
In the advertising industry, marketers strive for precise and relevant data regarding users. Our massive scale (more than 1.2B+ users reached monthly) allows us to gather a lot of data.
The Data Science team at Teads aims to provide other teams with a set of knowledge and tools enabling them to obtain a clear vision of their activities (i.e. aggregation of data on different dimensions, qualitative temporal analyzes, performance indicators, etc.) and to enable them to make strategic decisions in short, medium and long terms.
Teads is currently looking for a Data Scientist to strengthen this strategic team.
Your mission will be to :
- Contribute to the improvement of algorithms and prediction models allowing to analyze an increasing amount of data accurately on the activity of Teads.
- Contribute to the development of new models and new approaches to Machine Learning to better anticipate and forecast traffic on our platform.
- Take part in the team's brainstorming in a continuous improvement process of existing systems.
Examples of proposed internship topics:
Improvement of prediction models by learning set
The "ensemble learning" method consists in using the predictions of several models to answer the same question. The objective of the internship would be to apply this method to our advertising video viewing prediction model. The models to be assembled may differ by their features, their set learning (bagging), their weights (boosting)... Different ways of assembling the models will also be tested. The comparison of the new models will be done offline and the best method will be implemented in our Machine Learning library before being tested in
production via an ABtest.
Gradient boosted decision trees as selectors/feature builders
We use generalized linear regressions to predict viewing advertising videos, user clicks and other interactions with our advertisers. These methods require some manual adjustments, such as the bucketisation, cross features selection, etc. The Gradient Boosted Decision Trees (GBDT) are used, for example, at Facebook to make these adjustments automatically, see : Practical Lessons from Predicting Clicks on Ads at Facebook. The objective of the internship would be to study the use of GBDTs for the construction of our models. The comparison of the new models will be done offline and the best method will be implemented in our Machine Learning library before being tested in production via an ABtest.
What do we require?
- You are in a higher education course such as an Engineering School or University course equivalent.
- You studied statistics (i.e. statistical analysis, regression analysis...) and / or Artificial Intelligence (I.e. Data Mining, Machine Learning,...) during your courses
- You have knowledge of our technical environment (we use
mainly Scala & Spark).
- You speak fluently English.
Our Technical Stack:
Front : AngularJS, ReactJS, LESS, Grunt et Webpack
Back : Scala, Java, Python et NodeJS framework
Data : Spark, Cassandra, Kafka, Redis & MySQL
Outil : Git, JIRA, Jenkins et Confluence
Infra : AWS, GCP, Terraform et Docker
More information on our Engineering Blog
Teads, founded in 2011, is the inventor of outstream video advertising and the leading native video advertising marketplace. Publishers work with Teads to create brand new outstream video inventory, monetizing it through programmatic buying, their own sales force, or third parties including Teads Demand. Teads pioneers advertising experiences that respect the user, creating unprecedented levels of premium inventory which previously didn’t exist. Brands, agencies and trading desks can access this top-tier, premium inventory, available on the web and on mobile, in the Teads Marketplace. Through our managed services capabilities, the Teads team execute on their client’s behalf using its platform. Teads has a team of over 500 employees, 100 of which are in the innovation team, across 27 global offices.