Senior Research Scientist, TikTok - Trust and Safety

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

Builds machine learning and foundation models that detect internet abuse and fraud across TikTok's global platform. The Trust and Safety R&D team designs and deploys multimodal models for content safety, risk detection, compliance understanding and moderation automation, with the aim of protecting users and publishers at very large scale.

• Fraud and scams

What you'd do

• Lead design, training and deployment of LLM and VLM foundation models for content safety tasks across modalities.
• Improve zero-shot and few-shot generalization for multilingual, multimodal and low-resource safety scenarios.
• Build and keep up a labeling system for image and video safety data that produces training sets and evaluation metrics.
• Explore and implement RLHF strategies to align models with evolving safety policies and user intent.
• Optimize large-scale training pipelines and model serving with infra and data teams on centralized GPU resources.
• Work with product, policy and recommendation teams to put safety models into real moderation workflows.

What it asks for

• Solid experience in traditional machine learning.
• More than two years of research or development work on large models.
• Experience designing and deploying models across NLP and computer vision.
• Strong coding and debugging skills.
• Publications at top conferences such as NeurIPS, ICML, ICLR, CVPR or ICCV, or competitive AI challenge experience (preferred).

Terms in this role

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.