Research Scientist – Generative Recommendation & LLMs

TikTok; Seattle (WA); On-site; USD 154,000 - 301,000; Full time

Job summary

TikTok’s E-commerce Recommendation Foundation team is building the next‑generation recommendation intelligence, delivering a unified Foundation Model to support multi‑business and multi‑scenario systems—from retrieval to ranking to re‑ranking.

We pursue an event‑sequence‑driven generative paradigm, deeply integrating large language models (LLMs), multimodal understanding, reinforcement learning, and system optimization to advance intelligent agents and outcomes for users.

Qualifications

• MS/PhD in Computer Science or related field or equivalent industrial research experience.
• Solid theoretical foundation in machine learning, deep learning, or information retrieval.
• Proficiency in Python and familiarity with mainstream deep learning frameworks (e.g., PyTorch).
• Strong passion for intelligent recommendation systems and a self‑driven research mindset.

Responsibilities

• Build and optimize cross‑scenario shared Foundation Models to enable unified modeling and efficient inference.
• Advance the event‑sequence‑driven generative recommendation paradigm, integrating multimodal understanding and generative capabilities.
• Apply LLM technologies across retrieval, ranking, and re‑ranking stages; participate in model training, inference optimization, and system co‑design.
• Explore the integration of LLMs / VLMs with recommendation systems to develop adaptive and evolving intelligent recommenders.
• Research end‑to‑end generative recommendation and system optimization methods that balance efficiency and user experience.

Skills

Machine learning fundamentals

Python

PyTorch

Research mindset

Deep learning theory

Education

MS/PhD in Computer Science or related field

Tools

PyTorch

Job description

TikTok’s E-commerce Recommendation Foundation team is building the next‑generation recommendation intelligence, delivering a unified Foundation Model to support multi‑business and multi‑scenario systems—from retrieval to ranking to re‑ranking.

We pursue an event‑sequence‑driven generative paradigm, deeply integrating large language models (LLMs), multimodal understanding, reinforcement learning, and system optimization to advance intelligent agents and outcomes for users.