Research Scientist - LLM Applications for International E-commerce Scenarios - Global Frontier Tech Recruitment Program - 2027 Start (PhD)

discover products that match their interests. Leveraging cutting-edge risk control and platform governance technologies to ensure a secure shopping environment for all users. Constructing a comprehensive product knowledge graph and intelligent customer service systems to elevate the intelligence of every stage of the transaction process. Integrating machine learning with operations research algorithms to continuously optimize supply chain and logistics efficiency, thereby reducing operational costs. Providing merchants with intelligent business tools that empower them to enhance their operational efficiency and customer service experience. By joining us, you will collaborate with world-class product and technology teams, leveraging ByteDance's extensive traffic and data ecosystem to drive the deep integration and application of technology across various e-commerce scenarios.
Topic Content: In today's global e-commerce landscape, intelligent systems must operate across increasingly complex and dynamic business environments. Yet existing approaches still face limitations in long-horizon forecasting, cross-modal understanding, and holistic decision-making.This initiative is focused on building a next-generation foundational large model purpose-built for global e-commerce applications. The model will integrate key business dimensions-such as users, products, content, logistics, and inventory-into a unified representation to support deep, context-aware intelligence at scale.Building on this foundation, we are developing a modular, agent-driven architecture that enables advanced capabilities including task planning, tool use, multi-turn reasoning, and real-world environment interaction.
Topic Challenges: 1. Heterogeneous fusion and alignment: Unified modeling of user behavior sequences, product sales time-series signals, and multimodal product content to achieve deep semantic alignment across high-dimensional temporal data and multimodal representations. 2. Synergy between recommendation LLMs and world models: Reformulating the recommendation problem as a generative task of producing ranked item lists for users, and leveraging large model technologies to enable end-to-end recommendation modeling. 3. Tokenizer of recommendation items: Designing scalable tokenization mechanisms to encode billions of items into multimodal and semantically rich representations, supporting training and generation tasks. This includes pretraining over tens of terabytes of user behavior tokens, improving scaling law performance through optimized model architectures and training strategies, and reframing diverse recommendation tasks as post-training objectives. Recommendation modeling is further enhanced using RLVR-style approaches to maximize GMV and user experience. In addition, training and inference are co-optimized, with high-performance recommendation systems built on large-model inference frameworks such as SGLang. 4. Multimodal large models for e-commerce: Developing multilingual, multimodal large models tailored for e-commerce, achieving state-of-the-art (SOTA) performance across core e-commerce scenarios. Building on this foundation, we establish an e-commerce agent backbone to enable scalable deployment of agent applications across diverse use cases. 5. Agent evaluation, safety, and compliance: Establishing evaluation metrics and benchmarks aligned with real-world business scenarios to ensure the robustness, safety, and compliance of agent systems, particularly under highly constrained and adversarial conditions.
Topic Value: 1. Building a general-purpose multimodal foundation to enable power-law scaling through iterative advancements in models, data, and compute, thereby strengthening the infrastructure for scalable AI foundations. 2. Establishing a global e-commerce foundation model to drive GMV growth and user retention through generative recommendation, time-series large models, and agent-based systems, ultimately creating a high-leverage revenue engine.

Qualifications

Minimum Qualifications

1. Education & Foundation: Ph.D. in Computer Science, AI, Mathematics, or a related field, with a strong foundation in data structures, algorithms, and mathematical modeling.
2. AI/ML Expertise: Solid understanding and research experience in Deep Learning, NLP, CV, Reinforcement Learning, Generative Models, or Multimodal Learning.
3. Coding & Engineering: Proficient in major programming languages and machine learning frameworks (e.g., PyTorch, TensorFlow), combined with excellent problem-solving, self-learning, and teamwork skills.

Preferred Qualifications

1. Academic Excellence: Proven track record of publications in international AI/CS conferences or journals (e.g., NeurIPS, ICML, ICLR, CVPR, ACL, KDD, SIGIR, WWW) or top rankings in recognized algorithmic competitions.
2. Domain-Specific Impact: Hands-on experience in leading or participating in key projects related to Search, Advertising, Recommendation systems, or Large Language Models (LLMs).
3. Advanced Multimodal Applications: Specialized expertise in multimodal large models, particularly in long-text processing or applications within the film and television drama domains.