Research Scientist - Large Language Model Applications for Global E-commerce

Overview

We seek highly driven individuals to join our team in 2027, embarking on a career path that encourages innovation, complex problem solving, and expansive growth opportunities. Candidates must be available to start by the end of 2027, with availability and graduation date explicitly stated in their resume.

Team Profile

The Data-Global E-Commerce team operates as the technological core driving ByteDance's international e-commerce, focusing on cutting-edge innovations surrounding algorithms and big data. The mission is to facilitate easier access to quality products worldwide by developing advanced recommendation and search mechanisms, enforcing stringent risk controls and platform governance, building comprehensive product knowledge graphs and intelligent customer service systems, optimizing supply chain logistics via machine learning and operations research, and empowering merchants with intelligent operational tools. Joining the team means close collaboration with world-class product and technology experts, leveraging ByteDance’s extensive data and traffic ecosystems across diverse e-commerce settings.

Project Focus

The initiative aims to develop a next-generation foundational large language model specifically designed for global e-commerce. This unified model will represent crucial business aspects including users, products, content, logistics, and inventory, supporting context-aware intelligence at scale. A modular agent-driven architecture is central to facilitating capabilities such as task planning, tool utilization, multi-turn reasoning, and interaction with real-world environments.

Challenges to Address

• Integration and alignment of heterogeneous data including user sequences, product sales time-series, and multimodal content for deep semantic understanding.
• Reformulation of recommendation systems as generative tasks producing ranked item lists, leveraging large model technology for end-to-end solutions.
• Development of scalable tokenization to represent billions of items multimodally and semantically, supporting vast data pretraining and optimizing model architectures for enhanced performance.
• Creation of multilingual, multimodal large models optimized for e-commerce scenarios, underpinned by an agent backbone enabling scalable deployment of agent applications.
• Formulation of evaluation metrics ensuring agent robustness, safety, and compliance in real-world, constrained, and adversarial situations.

Impact and Value

• Building a versatile multimodal foundation to enable exponential scaling through iterative model, data, and compute improvements.
• Establishing a global e-commerce foundation model to enhance gross merchandise value (GMV) growth and user engagement via generative recommendation, time-series modeling, and agent-based systems, forming a scalable revenue engine.

Required Qualifications

• Ph.D. in Computer Science, Artificial Intelligence, Mathematics, or related disciplines, with strong knowledge of data structures, algorithms, and mathematical modeling.
• Substantial research knowledge and experience in Deep Learning, Natural Language Processing, Computer Vision, Reinforcement Learning, Generative Models, or Multimodal Learning.
• Strong programming skills in major languages and ML frameworks such as PyTorch or TensorFlow, paired with excellent problem-solving, self-directed learning, and teamwork capabilities.

Preferred Qualifications

• Demonstrated excellence evidenced by publications at top AI or CS conferences/journals like NeurIPS, ICML, ICLR, CVPR, ACL, KDD, SIGIR, WWW, or high rankings in algorithmic contests.
• Experience contributing to significant projects in Search, Advertising, Recommendation Systems, or Large Language Models.
• Expertise in advanced multimodal large models focusing on long-text processing or applications related to film and television drama domains.

About TikTok

TikTok is the premier platform for short mobile videos, committed to fostering creativity and joy globally. Headquartered in Los Angeles and Singapore, with offices worldwide, the company values diversity, innovation, and collaboration. TikTok nurtures a culture of learning, resilience, and an “Always Day 1” approach to create impactful solutions for users around the world.

Diversity and Inclusion

TikTok fosters a workplace embracing diverse skills, backgrounds, and perspectives to reflect the global communities it serves, championing inclusivity and celebrating varied voices in their mission to inspire and bring joy.