T

TikTok

Short-form video platform with ads

Applied Scientist Intern - Trust and Safety, Multimodal Foundation Model

Natural Language Processing & LLMsConfirmed live in the last 24 hours
Internship · Summer 2027San Jose, CA, USA$75 /hrIn-person

About the Job

Requirements

  • Currently pursuing a PhD in Computer Science, Data Science, Artificial Intelligence, or a related field.
  • Strong understanding of cutting-edge large language model research, including long context, multimodality, alignment research, and agent ecosystems.
  • Proficiency in programming languages such as Python, Rust, or C++, and a track record of working with deep learning frameworks such as PyTorch, DeepSpeed, Megatron, or vLLM.
  • Strong understanding of distributed computing frameworks, performance tuning, and verification for training, fine-tuning, and inference.

Responsibilities

  • Pursue research and implementation in multimodal moderation foundation models, including large-scale mixture-of-experts architecture training and routing optimization, cross-modal alignment and reasoning across text, images, video, and audio, unified understanding and generation, and synthetic data generation for moderation scenarios.
  • Develop reinforcement-learning-driven agentic moderation systems that improve multi-step decision-making, dynamically build moderation context, integrate tools, and support autonomous planning, tool collaboration, and interpretable closed-loop reasoning.
  • Address challenges in mixture-of-experts training stability, routing optimization, cross-modal token alignment, and unified architecture design for understanding and generation.
  • Train agent multi-step reasoning and tool-call strategies using GRPO or PPO while addressing sample efficiency and training stability.
  • Build context-engineering systems, including dynamic context assembly, MCP-based tool ecosystems, multi-source heterogeneous evidence fusion, and GraphRAG strategy retrieval.
  • Improve generalization across more than 200 languages and strategies, detect adversarially generated content, and design multidimensional reward signals for few-shot scenarios.
  • Attend onboarding during summer 2027; state availability clearly in the resume.

Desired Qualifications

  • Practical expertise implementing advanced large language model systems.
  • Familiarity with parameter-efficient fine-tuning, reinforcement learning, mixture of experts, chain of thought, or LangChain.
  • Ability to drive research projects from idea to implementation and produce tangible outcomes.
  • Published research papers or contributions to the large language model community.
  • Experience with inference tuning and acceleration, a deep understanding of GPUs or other artificial-intelligence accelerators, experience with large-scale AI networks, and experience with PyTorch 2.0 or similar technologies.
  • Experience evaluating AI systems and developing large language model applications or agents.

About the Company

TikTok

Short-form video platform with ads

TikTok is a short-form mobile video platform that lets people create, discover, and share quick videos. Users upload vertical clips, edit them with built-in tools, and browse a personalized feed powered by an algorithm that highlights relevant content. Advertisers can use in-feed ads, branded hashtags, and the Business Center to plan and measure campaigns, making it easy for brands to reach a global audience. Its goal is to inspire creativity and bring joy by helping people express themselves and by connecting brands with a wide audience.

Company size
10,001+
Company stage
Grant
Total funding
$740K
Headquarters
Santa Monica, CA
Founded
2016

Recently Posted Jobs

Explore more opportunities at TikTok

See more jobs