ByteDance Seed Division Restructuring: Multimodal Head Takes Over Embodied AI Business, Consolidating Robotics R&D Resources
Miles Bennett
ByteDance's Seed team has placed its robotics unit under multimodal head Zhou Chang, merging hardware, data, and model resources previously split across multiple teams. This means → ByteDance is shifting robotics from parallel tracks to a single chain of command, accelerating the division's pace.
What actually changed in this reshuffle?
Seed Robotics has reported to multimodal head Zhou Chang for over a month. Former head Li Hang now serves as an advisor focused on academic partnerships.
Zhou Chang now oversees multimodal interaction, world models, and embodied intelligence — three lines under one decision-maker.
Robotics R&D was previously split across teams handling hardware, data, and models separately, creating overlap. In plain terms = one group built the body, another trained the brain, a third gathered the data — now all three share a single resource pool.
How did Zhou Chang's scope grow so fast?
Zhou Chang joined Alibaba's DAMO Academy after completing his PhD in 2017 and worked on the Tongyi Qianwen large model. He moved to ByteDance in 2024 to lead multimodal interaction and world models.
Over the past year, the visual multimodal generation lead Yang Jianchao went on leave and the visual foundation-model research lead Feng Jiashi departed. Text-to-image model Seedream and text-to-video model Seedance were folded into Zhou Chang's remit.
This means → his expansion was not a single planned move but vacancy-driven and incremental — each departure pushed his boundary outward.
What has Seed Robotics been through?
The team originated as ByteDance's AI Lab robotics group. When AI Lab merged into Seed in 2025, the team entered the Seed structure. Li Hang reported directly to Seed head Wu Yonghui but was not heavily involved in day-to-day management.
Shortly after the merger, former research lead Kong Tao left to start his own company. Seed then began recruiting for multiple leadership roles — robotics product, engineering, and embodied-intelligence model heads — most of which remain unfilled.
In plain terms = the team went through "merger → key departures → rebuilding from scratch." Zhou Chang inherited a unit that needs to be reassembled.
What level is the new "embodied-intelligence tech lead"?
ByteDance is simultaneously hiring an embodied-intelligence tech lead to oversee the entire robotics roadmap, reporting to Zhou Chang.
The role is pegged at L8 — equivalent to Alibaba's P10–P11, a very senior technical rank.
Target candidates are CTOs of leading embodied-intelligence startups. This reflects ByteDance's intent to recruit proven operators rather than develop talent internally.
The bigger picture: why is everyone racing into robotics now?
On June 1, OpenAI CEO Sam Altman announced that OpenAI Robotics is hiring full-stack hardware, operations, systems, and ML engineers.
OpenAI's near-term goal is robots that help skilled workers build infrastructure; its long-term vision is a personal robot for everyone.
This means → model companies across the board are pouring resources into robotics. ByteDance's consolidation is not an isolated move — it is part of an industry-wide pivot from "large-model race" to "embodied deployment."
Content is for reference only, not financial advice.