Research Scientist (Trust and Safety - Vision Language Models/VLM)

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

Builds vision language models and other foundation models that feed TikTok's content moderation, search, recommendations and client AI work. The team's job is to turn foundation model world knowledge into features for downstream business use across the platform.

What you'd do

• Enhance VLMs with specialized features such as OCR and captioning for TikTok business applications
• Explore model architecture and inference-efficient design for scaled use in downstream TikTok business
• Work with cross-functional teams to plan and implement VLM projects for diverse purposes and vertical domains
• Carry research insights and impact from industry back to academia

What it asks for

• Doctorate in a computing, data, AI or closely related discipline
• Understanding of current LLM research such as long context, multimodality, alignment and agents
• Proficiency in Python, Rust or C++ and experience with deep learning frameworks such as PyTorch, DeepSpeed, Megatron or vLLM
• Experience with inference tuning and acceleration on GPUs or other AI accelerators (preferred)
• Published research or contributions to the LLM community (preferred)
• Experience evaluating AI systems and building LLM application or agent systems (preferred)
• Familiarity with distributed training and inference performance tuning, plus PEFT, RL, MoE, CoT or LangChain (preferred)

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.