Group Leader, Safety Model Operations (APAC) - AI Data Service Operations

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

Leads large operational teams in APAC that deliver AI moderation, labeling and safety data services, including human content moderation, annotation, red-teaming and evaluation. The team sits in Safety Model Operations and works to tie the quality of human operations to how well AI models perform on safety and alignment.

What you'd do

• Lead, coach and develop large operational teams and team leads delivering AI moderation and labeling work.
• Own operational KPIs and the link between operations output and downstream model impact.
• Run detailed reviews, complex root cause analyses and quality audits, tying findings to model-level outcomes.
• Monitor quality and compliance against project KPIs and flag trends, risks and guideline or rater pool gaps.
• Drive continuous improvement on annotation accuracy, inter-annotator agreement and cultural sensitivity.
• Work with product, engineering, policy and data science teams on experiments and tool or workflow changes.
• Build training programs on operations excellence and data-model feedback loops.

What it asks for

• Degree in a field such as business, management, operations, data science or similar.
• Understanding of how human operations affect AI training, feedback loops and safety alignment.
• Proficiency with SQL and data visualization tools such as Power BI for quality and IAA analysis.
• Solid grounding in machine learning workflows, data annotation and model evaluation.
• Experience with global rater pools and multilingual or cultural moderation challenges (preferred).
• Familiarity with RLHF, RLAIF, preference modeling and red-teaming (preferred).

Terms in this role

Red teaming.

Testing a system, often an AI model, by deliberately trying to make it fail or misbehave, so problems are found and fixed first.

RLHF.

Reinforcement learning from human feedback: training an AI model on people's ratings of its answers.

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.