CV/NLP/Multimodal LLM Research Scientist Graduate (Trust and Safety) - 2026 Start (PhD)

Develop computer vision and multimodal models to detect Trust & Safety violations, improving TikTok’s content understanding capabilities. Research and apply cutting-edge multimodal/vision large models and LLM techniques such as pre-training, zero-/few-shot learning, and hard case mining. Continuously optimize distributed model training frameworks to better support large-model training, collaborating with the Algorithm team to protect users and platform health.