Responsibilities
1. Create a reliable algorithm model streaming training and batch training architecture for Douyin Group's advertising business, including sample splicing, training strategy, unified sample storage, batch-stream integrated training framework, etc., to provide efficient and stable learning capabilities for Douyin Group's advertising algorithm model 2. Build an algorithm data warehouse and data lake that supports millions of advertising QPS requests and daily average PB-level data increments 3. Build an advertising platform and architecture to provide high-quality data for basic advertising targeting, algorithm model training and other scenarios 4. Build an advertising sample platform and architecture, define sample research and development standards and specifications, support massive business needs, and ensure sustainable iteration of the business.
Qualifications
1. Bachelor degree or above, computer and other related majors, with a solid computer foundation 2. Familiar with the Linux operating system and development environment, proficient in one or more programming languages such as Java/C++/Golang and an in-depth understanding of language features, with good data structure, algorithm foundation and system design capabilities 3. Good at communication, proactive work, strong sense of responsibility, and good teamwork skills 4. Have an in-depth understanding of streaming/batch computing systems, in-depth mastery of Flink/Spark implementation principles, TB-level Flink/Spark application and optimization experience, and experience in Flink/Spark source code reading will be preferred 5. Have an in-depth understanding of the platforming and productization of features and samples, and those with experience in machine learning system development will be preferred.
