Job description

About the Team TikTok’s Recommendation Foundation team builds shared foundation models for Global E-commerce. We are advancing event-sequence-driven generative recommendation with LLMs/VLMs, multimodal understanding, and reinforcement learning—moving beyond click prediction toward recommendation agents that understand user intent and connect people with products and content. We are looking for researchers and engineers with demonstrated LLM impact in industry or academia, or experience building and deploying generative recommendation systems at scale. You will shape foundation-model approaches for Global E-commerce and take them from research to production. Responsibilities - Own foundation models from research to production: define the modeling strategy and lead data, architecture, pre-training, mid-training, and post-training through deployment. - Advance LLM-native, event-sequence-driven recommendation: model user behavior and intent to improve retrieval, ranking, and end-to-end generative recommendation. - Build multimodal representations and semantic tokenizers that connect product, content, and behavioral signals and transfer across scenarios. - Deliver measurable online impact: establish rigorous evaluations and experiments, balancing recommendation quality and long-term user value with latency and inference cost. - Lead through hands-on execution: guide model and system design, align research and engineering partners, and mentor colleagues while staying close to code and experiments. Minimum Qualifications: - MS/PhD in Computer Science, related technical field or equivalent industrial research experience. - Demonstrated impact in LLMs through industry systems or academic research, or hands-on experience building and deploying large-scale generative recommendation systems. - Strong machine learning and deep learning fundamentals, with expertise in LLMs, foundation models, or generative recommendation - Ability to lead complex technical work from problem definition through evaluation and deployment, and collaborate across research and engineering. Preferred Qualifications： - Experience setting research roadmaps, leading shared modeling platforms or cross-team initiatives, and mentoring researchers and engineers. - Experience with foundation-model training or post-training, reinforcement learning, or preference optimization. - Expertise in generative retrieval, semantic tokenization, multimodal understanding, long-sequence user modeling, or efficient training and inference. - Evidence of impact through production launches, reusable technical contributions, or publications at leading ML/NLP venues.

Requirements

Must-haves

About the Team
Own foundation models from research to production: define the modeling strategy and lead data, architecture, pre-training, mid-training, and post-training through deployment.
Advance LLM-native, event-sequence-driven recommendation: model user behavior and intent to improve retrieval, ranking, and end-to-end generative recommendation.
Build multimodal representations and semantic tokenizers that connect product, content, and behavioral signals and transfer across scenarios.
Deliver measurable online impact: establish rigorous evaluations and experiments, balancing recommendation quality and long-term user value with latency and inference cost.
Lead through hands-on execution: guide model and system design, align research and engineering partners, and mentor colleagues while staying close to code and experiments.
MS/PhD in Computer Science, related technical field or equivalent industrial research experience.
Demonstrated impact in LLMs through industry systems or academic research, or hands-on experience building and deploying large-scale generative recommendation systems.
Strong machine learning and deep learning fundamentals, with expertise in LLMs, foundation models, or generative recommendation
Ability to lead complex technical work from problem definition through evaluation and deployment, and collaborate across research and engineering.
Experience setting research roadmaps, leading shared modeling platforms or cross-team initiatives, and mentoring researchers and engineers.
Experience with foundation-model training or post-training, reinforcement learning, or preference optimization.
Expertise in generative retrieval, semantic tokenization, multimodal understanding, long-sequence user modeling, or efficient training and inference.
Evidence of impact through production launches, reusable technical contributions, or publications at leading ML/NLP venues.

Nice-to-haves

Preferred Qualifications：

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

llmmachine learningml

Platform: tiktok
Effort: medium