ByteDance
Research Scientist Graduate (Seed Model - Multimodal Interaction & World Model) - 2026 Start (PhD)
San Jose, California · Full-time · Posted Oct 28, 2026
- Education
- PhD in a listed or related discipline; completing or recently completed.
About the job
Conduct research and engineering on multimodal models, improving their understanding, reasoning, performance, and efficiency through scaling-law analysis and systematic experiments.
What you'll do
- Advance multimodal understanding and reasoning capabilities through research and engineering.
- Explore methods to improve model performance and efficiency.
- Establish scaling laws and conduct systematic ablations to produce transferable conclusions.
What they ask for
- Completing or recently completed a PhD in Computer Science, Electrical Engineering, Electrical and Computer Engineering, Physics, Mathematics, or a related discipline.
- Publications in accredited venues such as CVPR, ECCV, ICCV, NeurIPS, ICLR, ICML, or other leading AI and ML conferences.
- Strong research background in at least one of reinforcement learning, multimodal learning, video understanding, or vision-language modeling.
Nice to have
- Expertise in Transformers (Dense and MoE) and familiarity with scaling Transformers on GPUs or TPUs.
- Hands-on experience with PyTorch/JAX and distributed training frameworks, or familiarity with state-of-the-art multimodal training-data preparation techniques.
Tools
Transformers (Dense and MoE), GPUs, TPUs, PyTorch/JAX
AI focus
Research and engineering for multimodal understanding, interaction, and reasoning models, including optimization of model performance and efficiency.
Model-written summary · the original posting is authoritative