Applied Scientist Intern - Trust and Safety (Multimodal Foundation Model) - Global Frontier Tech Recruitment Program - 2027 Start (PhD)

TikTok
San Jose
Workplace: OnsiteInternshipFunction: Research & Scientific (R&D)Skills: ["Problem-solving","Creative mindset"]

Work on TikTok Trust & Safety to build multimodal moderation foundation models and agentic moderation systems. Research and implement large-scale Mixture-of-Experts (MoE) training/routing, cross-modal alignment, and unified understanding/generation. Extend this with reinforcement learning for multi-step agent decision-making, tool ecosystems (MCP), and retrieval (GraphRAG), improving robustness across many languages and adversarial AIGC content.

Loading

Loading job details...

Preparing the role view and application actions.

FursaFursa
TikTok
TikTok
1 month ago

Applied Scientist Intern - Trust and Safety (Multimodal Foundation Model) - Global Frontier Tech Recruitment Program - 2027 Start (PhD)

✓ Verified Job

Canonical indexed version, validated from employer's careers page.

Source: Company careers pageValidated by: Fursa AI
Last checked: 23 hours agoStatus: Live

Job Summary

Work on TikTok Trust & Safety to build multimodal moderation foundation models and agentic moderation systems. Research and implement large-scale Mixture-of-Experts (MoE) training/routing, cross-modal alignment, and unified understanding/generation. Extend this with reinforcement learning for multi-step agent decision-making, tool ecosystems (MCP), and retrieval (GraphRAG), improving robustness across many languages and adversarial AIGC content.
Location: San Jose
Workplace: Onsite
Employment Type: Internship
Job Function: Research & Scientific (R&D)
Seniority: Intern level

Key Responsibilities

  • •Research and develop multimodal moderation foundation models, including large-scale MoE architecture training/routing optimization and cross-modal alignment.
  • •Build and improve synthetic data generation for moderation scenarios using approaches like self-play and adversarial augmentation.
  • •Develop agentic moderation systems using reinforcement learning to enable multi-step decision-making and tool-call strategies.
  • •Design context engineering and tool collaboration via MCP-based tool ecosystems and multi-source evidence fusion with GraphRAG retrieval.
  • •Evaluate and improve robustness through generalization across many languages/strategies and adversarial detection of AIGC content.

Key Requirements

  • •Currently pursuing a PhD in Computer Science, Data Science, Artificial Intelligence, or a related field.
  • •Strong understanding of cutting-edge LLM research (e.g., long context, multi modality, alignment research, agent ecosystem) with practical implementation experience as a plus.
  • •Proficiency in Python, Rust, or C++ and experience implementing deep learning systems with frameworks such as PyTorch, DeepSpeed, Megatron, or vLLM.
  • •Strong understanding of distributed computing and performance tuning/verification for training, finetuning, and inference; PEFT, RL, MoE, CoT, or LangChain familiarity is a plus.
  • •Ability to drive research projects from idea to implementation and produce tangible outcomes; published research contributions are a plus.
Skills:Problem-solvingCreative mindset
Tech Stack:PythonRustC++PyTorchDeepSpeedMegatronVLLMPEFTRLMoECoTLangChainGRPOPPOMCPGraphRAGGPU

Company Brief

TikTok
Short-form video platform that lets users create, share, and discover entertainment content through algorithmic recommendations. It also offers advertising and creator tools for brands, influencers, and businesses.
Industry: Digital Media
Company Size: Enterprise (1,001+ employees)
Growth: Scaleup
Headquarters: Singapore, Singapore
Founded: 2016
WebsiteLinkedIn