Our Challenges

Serving relevant ads at scale is actually quite a challenge. From running code in very heterogeneous contexts to handling large data streams and computing online predictions.

Run everywhere: The web is an untamed and ever-changing beast. Our first challenge is related to the user experience. We build ad formats that respect the user and must run in a wide range of contexts (device, OS, browser, websites, apps, etc.).

Deal with large streams of data: With billions of events collected, stored and processed each day, we have to build large scale and efficient data pipelines in order to fuel analytics dashboards and feed our prediction models with fresh and clean data.

Online predictions: Machine Learning is at the core of many of our innovations. From user behavior predictions, fraud detection to ad delivery pacing and infrastructure resources optimization. All these predictions need to be computed in a few milliseconds.

Computing efficiency: We handle millions of transactions per second and need to compute auction winners as fast as possible, considering many rules related to both campaigns and user contexts.

Distributed cloud infrastructure: Based on an active-active redundant architecture our cloud infrastructure is built to scale and handle significant traffic variations while staying highly available.

Keep up the pace: Our industry evolves quite rapidly, our architecture design and tooling must be in line with that so that our organization & product offering can scale.

A few numbers

0 Bn
Users/month
0 Bn
Analytics events/day
0 M
Transactions/sec
0 K
AWS instances

Our evolving stack

Scroll to Top