TikTok's US Core Recommendation Team aims to elevate personalized content discovery and user experiences by improving recommendation precision, user involvement, and scalability for hundreds of millions of users.
Requirements
- Currently pursuing a PhD degree in Computer Science, Electrical Engineering, Statistics, or a related field, with a focus on recommendation systems, natural language processing, or multimodal learning.
- Strong theoretical foundation and hands-on research experience in relevant areas.
- Proficiency in Python and familiarity with ML frameworks such as PyTorch or TensorFlow.
- Solid foundation in data structures, algorithms, and analytical and problem solving skills.
- First-author publications in top-tier conferences such as NeurIPS, ICML, ACL, CVPR, or KDD.
- Experience with large-scale machine learning systems or applied research in industry.
- Prior work or research integrating LLMs or multimodal models into real-world applications.
Responsibilities
- End-to-End Generative Large Recommendation Systems: We are committed to reimagining the traditional recommendation pipelines. You will explore novel architectures, algorithms, and optimization strategies to break through the limitations of existing systems. By challenging the status quo, you will strive to build more efficient, scalable, and generative recommendation frameworks.
- Ultra-Long Sequence Modeling of User Lifecycle Behavior: Understanding user behavior over an extended period is crucial for providing long-term personalized recommendations. You will focus on modeling the ultra-long sequences of user interactions throughout their lifecycle on TikTok.
- Integrating LLM and Multimodal Technologies for Recommendation: With the abundance of multimodal content (text, image, video, audio) on TikTok, integrating LLM and multimodal technologies into recommendation systems is essential. You will work on leveraging the power of LLMs to understand and process information, and combine it with other multimodal data to enable seamless multimodal-recommendation fusion.
- Posttraining & RL: Exploration of posttraining methods to better align large generative models with business and feed quality needs. Conduct original research on applying RL (e.g., bandit models, policy optimization, offline RL) to recommendation problems (such as diversity & multiobjective fusion problems)
- Conduct in-depth research and development in the aforementioned groundbreaking directions, designing and implementing innovative algorithms to enhance recommendation performance and accuracy.
- Analyze large-scale user behavior data and content data to gain insights and drive model improvements.
- Participate in the deployment and evaluation of the developed recommendation systems in real-world scenarios, ensuring their practical effectiveness.
Other
- We are looking for talented individuals to join our team in 2026.
- We will prioritize candidates who are able to commit to work with the team for 12 weeks.
- Please state your availability for the internship in your resume.
- Applications will be reviewed on a rolling basis. We encourage you to apply early.
- Collaborate with cross-disciplinary teams, including infrastructure engineers, PMO, and researchers, to create advanced systems that improve recommendation relevance, diversity, and user engagement.
