Machine Learning Researcher - Search

Team Overview

The Search Team supports key products including Douyin, Toutiao, Xigua Video, and business verticals like E-commerce and Local Services by innovating search algorithms and architectures. Our work leverages state-of-the-art machine learning techniques for end-to-end modeling and system performance optimization across memory, disk, indexing, recall, and ranking algorithms, providing extensive growth opportunities.

Core Responsibilities

• Advancing NLP methods from fundamental tasks such as word segmentation and Named Entity Recognition (NER) to sophisticated functions including text and multimodal pre-training, query analysis, and relevance modeling using deep learning.
• Developing cross-modal matching technology integrating computer vision and natural language processing for semantic understanding and retrieval in multimodal video search.
• Applying large-scale streaming machine learning to personalize search recommendations and enhance user needs comprehension.
• Conducting large-scale data architecture research spanning offline computing, distributed system optimization, and building high-availability, low-latency online services.
• Innovating search recommendation algorithms using ultra-large-scale machine learning to maintain industry leadership.

Key Focus Areas

• Integrating large models with ranking algorithms to refine personalized rankings and enrich user experience.
• Developing generative retrieval algorithms capable of handling candidate libraries containing tens to hundreds of billions of entries.
• Utilizing large language models (LLMs) to boost search satisfaction for complex or ambiguous queries.
• Constructing highly efficient batch-stream integrated retrieval and computing systems to optimize resource utilization.

Challenges Addressed

• Improving personalized ranking by leveraging multimodal data beyond traditional algorithms.
• Overcoming limitations of current discriminative models for ultra-large-scale retrieval by innovating next-generation retrieval algorithms.
• Enhancing understanding of complex user queries that involve polysemy and long-form semantics to increase search satisfaction.
• Boosting resource usage efficiency in storage-computation separated search system architectures while maintaining performance.
• Incorporating large model technologies to upgrade semantic understanding, retrieval effectiveness, and resource management in search systems.

Research Directions

• Developing large-scale models for ranking purposes.
• Exploring generative retrieval and fusion approaches for cross-modal data.
• Leveraging LLMs to understand complex queries better.
• Designing high-performance compute and storage architectures supporting large-scale applications.

Candidate Qualifications

• PhD degree preferred, ideally in AI, computer science, natural language processing, computer vision or adjacent domains.
• Track record of publications at flagship AI conferences or demonstrated deep research expertise.
• Strong foundation in machine learning and deep learning algorithms with proficient coding skills in C/C++ or Python.
• Driven, confident, technology passionate, and eager to push boundaries.
• Effective communication skills and ability to collaborate within diverse teams.

Company & Culture

TikTok, globally recognized for short-form mobile video, is dedicated to inspiring creativity and spreading joy. With headquarters in Los Angeles and Singapore plus worldwide offices, TikTok fosters innovation through curiosity, humility, and resilience. The company promotes an inclusive environment valuing diverse experiences and viewpoints, striving to make an impactful difference for users and teams alike.