Research Scientist Intern (Data-TnS-Algo-Foundations & Intelligence Service) - 2027 Start (PhD)
Core
Researching core development and research for TikTok's foundation-model team to push boundaries of LLM and VLM capabilities with trustworthiness and safety built in.
Role type
PhD Research Scientist Intern (LLM/VLM Safety & Alignment)
Builds
Foundation models (LLMs and VLMs) for product teams and customers
Domain
Artificial Intelligence, Large Language Models, Vision-Language Models
Deliverable
production ML models
Required skills
Deep learning frameworks (PyTorch, TensorFlow), LLMs, AI Safety, Computer Vision, Multimodality, Pretraining & Continued Pretraining, Evaluation & Measurement, Reinforcement Learning (RLHF/RLAIF/DPO), Agent workflows
Preferred skills
Publications in top AI or Security venues (NeurIPS, ICLR, ICML, CVPR, etc.)
Responsibilities
Develop approaches to improve general capability and safety for LLMs/VLMs via pretraining/CPT; Build advanced evaluation systems to study emerging LLM/VLM skills and safety behaviors; Develop advanced reinforcement-learning strategies and investigate balancing pretraining and post-training; Probe new failure modes in scaffolded environments (agents, workflows, tool use) and translate insights into robust mitigations
Seniority
Intern, PhD candidate
