Multimodal Trust & Safety Scientist — Foundation Models

TikTok Inc.; Seattle (WA); On-site; USD 207,000 - 368,000; Full time

Job summary

TikTok Inc. is seeking graduates for advanced research in Trust & Safety, focusing on multimodal moderation foundations and agentic systems. The role involves building large-scale models, cross-modal alignment, and reinforcement learning-driven decision-making to improve content safety globally.

Strong background in LLMs, Python/C++, and DL frameworks is expected, with a passion for scalable AI and multilingual moderation strategies across platforms.

Qualifications

• PhD in Computer Science, Data Science, AI, or related field.
• Strong understanding of cutting-edge LLM research and practical implementation.
• Proficiency in Python and C++, with experience in DL frameworks.
• Familiarity with PEFT, RL, MoE, CoT or Langchain is a plus.

Responsibilities

• Research and develop multimodal safety foundation models for moderation.
• Advance RL-driven agentic decision-making and tool collaboration.
• Research scalability and robustness across languages and modalities.

Skills

Python

C++

PyTorch

Distributed computing

PEFT

Education

PhD in CS/AI

Tools

PyTorch

DeepSpeed

Megatron-LM

VLLM

Job description

TikTok Inc. is seeking graduates for advanced research in Trust & Safety, focusing on multimodal moderation foundations and agentic systems. The role involves building large-scale models, cross-modal alignment, and reinforcement learning-driven decision-making to improve content safety globally.

Strong background in LLMs, Python/C++, and DL frameworks is expected, with a passion for scalable AI and multilingual moderation strategies across platforms.