About the job
TikTok Commerce Ads Ranking team is dedicated to model optimization in the ads delivery system. Our goal is to improve monetization efficiency of TikTok Commerce Ads products through model optimization in full funnel. Besides, our team aims to design and implement universal modeling solutions, and solve the long-standing problems of ranking algorithms. At TikTok Commerce Ads Ranking team, you can optimize model in recall / rough sort / fine sort, and explore innovative algorithms to break through the ceiling of ads performance.
Responsibilities
1. Optimizing model in ads delivery system: feature engineering, model structure, auto crossing, ads cold start, modeling delayed feedback, multi-task learning, sequence modeling
2. Algorithm and system co-design: retrieval algorithm, sample mining, long sequence
3. Exploring large-scale distributed training framework: GPU tuning, feature processing, synchronous training
Improving ads delivery efficiency in privacy-preserving environments
4. LRM, the next generation rec system using LLM learning paradigm: entity understanding, end2end Generative Recommendation, sequence only based Recommendation, Mixture of Export
Qualifications
Minimum
1. Have excellent coding skills, be familiar with C++/Python, and have a solid foundation in data structures and algorithms.
2. Solid foundation in machine learning, familiar with common algorithm models such as FM, LR, GBDT, DNN, Wide&Deep, etc.
3. Excellent learning ability and good teamwork spirit. And excellent analytical and problem-solving skills.
Papers in conferences such as KDD, NeurIPS, SIGIR etc, or experience in data mining/machine learning related competitions are a good fit.
4. Bachelor degree or above in computer science or related majors
Preferred
1. Papers in conferences such as KDD, NeurIPS, SIGIR etc, or experience in data mining/machine learning related competitions are a good fit.
2. Bachelor degree or above in computer science or related majors, work experience in advertising/recommendation/search ranking are preferred.