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

TikTok
San Jose
Workplace: OnsiteFull timeFunction: Data Science & Machine LearningEducation: phdSkills: ["Problem-solving","Creative mindset","Research execution","Cross-functional collaboration"]

Join the Foundations and Intelligence Service R&D team to build and improve LLM/VLM/Omni foundation models. You will enhance VLM performance with features like OCR and captioning, explore inference-efficient model architectures, and collaborate cross-functionally to apply VLMs to TikTok trust-and-safety use cases. The role also emphasizes evaluation, pre/post-training data processing, reinforcement learning-based alignment, and translating research insights to real-world outcomes and academia.

Loading

Loading job details...

Preparing the role view and application actions.

FursaFursa
TikTok
TikTok
1 month ago

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

✓ Verified Job

Canonical indexed version, validated from employer's careers page.

Source: Company careers pageValidated by: Fursa AI
Last checked: 1 minute agoStatus: Live

Job Summary

Join the Foundations and Intelligence Service R&D team to build and improve LLM/VLM/Omni foundation models. You will enhance VLM performance with features like OCR and captioning, explore inference-efficient model architectures, and collaborate cross-functionally to apply VLMs to TikTok trust-and-safety use cases. The role also emphasizes evaluation, pre/post-training data processing, reinforcement learning-based alignment, and translating research insights to real-world outcomes and academia.
Location: San Jose
Workplace: Onsite
Employment Type: Full time
Job Function: Data Science & Machine Learning
Seniority: Graduate level

Key Responsibilities

  • •Enhance VLMs with specialized features like OCR and captioning to optimize TikTok business applications.
  • •Explore model architectures and inference-efficient design for scalable downstream TikTok applications.
  • •Plan and implement projects harnessing VLMs across diverse purposes and vertical domains with cross-functional teams.
  • •Contribute to evaluations and data processing for pre-training and post-training.
  • •Support reinforcement learning-based alignment and efficient training and inference work.

Key Requirements

  • •Completing or recently completed a PhD in Computer Science, Data Science, Artificial Intelligence, or a related field.
  • •Proficiency in Python, Rust, or C++ and experience with deep learning frameworks such as PyTorch, DeepSpeed, Megatron, or vLLM.
  • •Ability to work on LLM/VLM pretraining and application work, including evaluations and data processing for pre- and post-training.
  • •Experience with reinforcement learning-based alignment and efficient training and inference.
  • •Strong foundation in cutting-edge LLM research such as long context, multi-modality, and alignment.
Experience:Deep learningLLMVLMFoundation models
Education:PhD / Doctorate
Skills:Problem-solvingCreative mindsetResearch executionCross-functional collaboration
Tech Stack:PythonRustC++PyTorchDeepSpeedMegatronVLLMPyTorch 2.0GPUDistributed computingPEFTRLMoECoTLangChain

Company Brief

TikTok
Short-form video platform that lets users create, share, and discover entertainment content through algorithmic recommendations. It also offers advertising and creator tools for brands, influencers, and businesses.
Industry: Digital Media
Company Size: Enterprise (1,001+ employees)
Growth: Scaleup
Headquarters: Singapore, Singapore
Founded: 2016
WebsiteLinkedIn