Research Scientist (Graduate) – LLMs, Recs & Multimodal

ByteDance; San Jose (CA); On-site; USD 140,000 - 220,000; Full time

Job summary

ByteDance in San Jose seeks researchers to push the boundaries of recommendation systems and large language model research, focusing on multimodal models and scaling to hundreds of billions of parameters.

You will collaborate with engineering teams to apply deep learning in TikTok products, spanning recommendation, advertising, multimodal understanding, and other compute-intensive domains, and explore new architectures with strong team collaboration.

Qualifications

• Undergraduate/Master's degree in CS or related discipline.
• Track record of leading high-impact projects or publications in recommender systems, advertising, or large language model research is a strong plus.
• Solid foundations in machine learning and NLP, with strong research and problem-solving capabilities; publications at top-tier venues such as ACL, NeurIPS, ICML, ICLR, or CVPR are highly preferred.

Responsibilities

• Conduct cutting-edge research on the fusion of recommendation systems and LLMs/multimodal models, with a focus on areas including LLM4Rec, cross-platform and cross-scenario Foundation Models, hundred-billion-parameter scale models, and multimodal large models.
• Drive the application and optimization of deep learning algorithms within TikTok products, spanning recommendation, advertising, multimodal understanding, and other compute-intensive domains.
• Collaborate closely with engineering teams to explore and validate novel algorithms under new architectures.

Skills

Machine learning

NLP

Publications

Education

Undergraduate/Master's degree in CS

Job description

ByteDance in San Jose seeks researchers to push the boundaries of recommendation systems and large language model research, focusing on multimodal models and scaling to hundreds of billions of parameters.

You will collaborate with engineering teams to apply deep learning in TikTok products, spanning recommendation, advertising, multimodal understanding, and other compute-intensive domains, and explore new architectures with strong team collaboration.