AI Model Reinforcement Japanese Speaking Graduate (Safety Model Operations) - 2027 Start

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

Works on training and evaluation data for the machine learning models behind TikTok's automated Trust & Safety systems, which identify harmful content across regions. The Safety Model Operations team builds and maintains those models and the operational processes around them, and this role sits in its data quality work for large-model training and evaluation, with liaison to the Japanese market.

What you'd do

• Help shape and keep updating the quality benchmarks, decision rules, and sign-off criteria used in model training and evaluation.
• Review data outputs against quality standards, flag issues, and close the loop on corrections.
• Assess complex, ambiguous, and edge cases, and turn findings into reusable decision rules and knowledge assets.
• Work with production, product, algorithm, and strategy teams to align standards across the standards, production, acceptance, and feedback loop.

What it asks for

• A bachelor's degree is required; majors such as law, philosophy, economics, journalism, languages, or social sciences are preferred.
• Strong English across listening, speaking, reading, and writing, including reliable understanding of English video and text; IELTS 7.5, TOEFL 105, or equivalent.
• Japanese proficiency required for daily liaison with the local team.
• Familiarity with Excel and common data-processing tools (preferred).
• Hands-on use of mainstream foundation-model products and a basic sense of AI capability limits (preferred).
• Clear structured reasoning and strong written and verbal communication; detail-oriented.

Worth knowing

• Must commit to an onboarding date by the end of the year.

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.