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

Written by the T&S Network from TikTok's posting.

Builds machine learning models and systems that moderate user content across TikTok's platform, using large multimodal foundation models and agentic systems to reduce dependence on human reviewers and external APIs. The team works on multilingual and multimodal content, evolving policies, and adversarial AIGC attacks that the traditional classification approach struggles to handle.

What you'd do

• Research MoE architecture training and routing optimization for a multimodal moderation foundation model.
• Develop cross-modal alignment and reasoning across text, image, video and audio for moderation.
• Build synthetic data generation, including self-play and adversarial augmentation, for moderation scenarios.
• Apply RL methods such as GRPO and PPO to train multi-step agent decision-making and tool-call strategies.
• Design context assembly, tool ecosystem integration and multi-source evidence fusion for an agentic moderation system.
• Work on generalization across 200+ languages and adversarial detection of AIGC content.

What it asks for

• Enrolled in a PhD program in computer science, data science, AI or a closely related discipline.
• Proficiency in Python, Rust or C++ and experience with deep learning frameworks such as PyTorch, DeepSpeed, Megatron or vLLM.
• Strong understanding of distributed computing, performance tuning and verification for training, finetuning and inference.
• Published research or contributions to the LLM community (preferred).
• Experience with inference tuning and acceleration, including GPU or other AI accelerator work (preferred).
• Experience evaluating AI systems and developing LLM or agent applications (preferred).

Our best summary of the posting, and it can be wrong. The employer's posting is what counts: read it in full on TikToknorth_east before you apply. If we got something wrong, report this listing.