Research Scientist Graduate (Multimodal Interaction and World Model) - 2026 Start (PhD)

Work on ByteDance’s Seed Multimodal Interaction and World Model research team to advance multimodal understanding and reasoning. Drive research and engineering to improve models, explore ideas that improve performance and efficiency, and build scaling laws with systematic ablations. Contribute within a lab-style environment spanning multimodal learning, reinforcement learning, video understanding, and vision-language modeling, using Transformers and distributed GPU/TPU training.