Search Safety PM Project Intern (TikTok Safety Product) - 2026 Start

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

Builds and maintains training data for LLMs used in TikTok's content safety systems, supporting review efficiency, risk control and user trust on search. The Safety Product team sits on short-term project work that feeds model training and evaluation.

• AI misuse and jailbreaks

What you'd do

• Help build and daily maintain LLM training datasets and knowledge bases for accuracy and contextual completeness.
• Help design and iterate labeling systems and taxonomies for supervised and reinforcement learning workflows.
• Generate, review and polish large volumes of LLM training data against unified standards and evaluation criteria.
• Draft and optimize prompts for training and evaluation, then analyze model outputs for deficiencies, biases and failure patterns.
• Turn complex multi-domain content into standardized structured training inputs by extracting core intent and key information.
• Check datasets, labels and prompts for consistency, then tighten production standards using model feedback.

What it asks for

• Currently enrolled as an undergraduate university student.
• Commitment of at least 3 months, ideally a minimum of 4 working days per week.
• Strong ability to interpret and structure complex text with attention to detail and nuance.
• Write clear, precise, logically organized English at high volume.
• Strong interest in LLMs and fast learning of AI concepts, tools and workflows.
• Study in humanities, social sciences, languages, international relations or related fields, current or recent (preferred).
• Past work in annotation, labeling, research support or AI operations (preferred).

Worth knowing

• Fixed-term internship with applications reviewed on a rolling basis.

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.