Research Scientist in Multimodal Interaction and World Model - Seed - Graduates - 2027 Start (PhD)

About the team

The Seed Multimodal Interaction and World Model team is dedicated to developing models that have human-level multimodal understanding and interaction capabilities. The team is working to advance the exploration and development of multimodal assistant products.

Responsibilities

• Develop multimodal foundation models integrating vision, language, audio, and environment signals.
• Design and optimize world models for reasoning, planning, and interaction.
• Build training pipelines including data curation, alignment, and reinforcement learning.
• Improve agent capabilities such as perception, memory, decision-making, and tool use.
• Explore next-generation interaction paradigms between humans and intelligent systems.

Minimum Qualifications:

• Currently pursuing a PhD in computer science, mathematics, engineering, or a related field, with an expected graduation date in 2027 and the ability to commit to an onboarding date by the end of 2027.
• Excellent coding ability, data structures, and fundamental algorithm skills, proficient in C/C++ or Python, etc.
• Experience in multimodal learning, reinforcement learning, or agent systems.
• Familiarity with large-scale model training or simulation environments.

Preferred Qualifications:

• Strong research track record in relevant areas.
• Strong problem-solving and collaboration skills.