With the growth of Xiaomi’s advertising business and the increasing number of algorithm experiments, how to improve the experimental efficiency of algorithm students has become a problem we are faced with. In this context, we have built a machine learning platform, which helps us to solve the feature data sharing in the algorithm process and solve a series of problems such as the inconsistent results of offline and online experiments. In this speech, we will share with you how Xiaomi uses Apache components such as Pegasus and Flink to build xiaomi machine learning platform.

Speakers:

Fei Huang: millet, Responsible person of Commercial Platform Technology Department, Internet Business Department, Commercial platform, head of the department of technology, the Internet sales department is responsible for the app store, browser, security center, union, and other business to liquidate r&d, lead the technical team to complete the advertising engine, on the platform, algorithms, machine learning framework, and other construction of distributed system and frame, support billions every day advertising request, is the escort advertising revenue.