Wen Shilei, Head of ByteDance's Commercialization GenAI R&D, Resigns and Joins Embodied AI Company Octopus Dynamics
Wen Shilei, head of ByteDance's commercialization GenAI R&D, has completed his resignation and joined embodied AI company Octopus Dynamics as a partner. Previously, within four months, the team's head Peng Bingyue and China head Yuan Zehuan resigned in succession; the remaining personnel have been reassigned, and the independent organizational structure has effectively come to an end. Octopus Dynamics was founded in January 2026 and has completed a seed round of nearly $50 million, with cumulative funding of nearly 1 billion yuan.
Body · AI translation
Recently, Wen Shilei, the head of GenAI R&D for ByteDance's commercialization business, was reported to have completed all departure procedures, with his next stop being the embodied intelligence company Octopus Dynamics, which he is joining as a partner.
Wen Shilei's technical foundation was laid during his ten years at Baidu.
He earned both his bachelor's and master's degrees from Huazhong University of Science and Technology, joined Baidu through campus recruitment in 2011, started as a search R&D engineer, and gradually grew into a key figure in Baidu's visual technology system. During his time at Baidu, he served successively as head of image and video technology in the Visual Technology Department, chairman of the Baidu Video Understanding Technology Committee, and principal architect of the Smart City Business Unit, reaching job grade T9, one of the highest grades in Baidu's technical track.
He led his team to more than 20 championships in international image and video recognition competitions, published more than 30 papers in top conferences and journals such as CVPR, ICCV, ECCV, and IEEE TPAMI, and twice won Baidu's highest award. At a stage when visual AI was moving from the laboratory to industrial-grade applications, Wen Shilei was one of the few technical leaders inside Baidu with both frontier research judgment and large-scale engineering deployment experience.
After moving to ByteDance in 2021, Wen Shilei's trajectory of responsibilities closely overlapped with several shifts in ByteDance's AI strategy.
He initially served as head of XR technology for intelligent creation, leading the large-scale deployment of technologies such as digital humans and AR/VR across businesses including Douyin, PICO, and Volcano Engine. As ByteDance integrated AI capabilities for commercialization scenarios into an independent R&D unit, Wen Shilei moved to become head of R&D for that GenAI team, reporting to Peng Bingyue, head of the commercialization business line, and taking full charge of engineering architecture design and technology productization for generative AI in high-value scenarios such as advertising delivery and intelligent marketing. In the words of people close to the team, his role was to "embed" generative AI into ByteDance's core advertising machine so that model capabilities would directly act on commercialization efficiency.
But the organizational vehicle for this monetization path underwent a complete dismantling over the past four months.
In July 2026, Peng Bingyue, head of the commercialization GenAI department, left and subsequently joined Google; in the same month, Yuan Zehuan, head of the China region, left ByteDance to start a business in the world model direction. In October, Wen Shilei completed his departure procedures, becoming the third core leader to leave. With his departure, the remaining personnel of the team have been transferred to other departments, and the original independent organizational setup has effectively come to an end within ByteDance.
The direction of this shift can be traced in ByteDance's recent organizational moves.
On July 30, 2026, ByteDance launched an adjustment to the organizational structure of its AI business, integrating the Feishu GTM team with the corresponding team at Volcano Engine to establish a "Creativity Service Platform" To B GTM organization, which is responsible overall for the marketing, sales, and customer service of ByteDance's cloud services such as MaaS and SaaS.
ByteDance CEO Liang Rubo said at the Volcano Engine conference in June that "climbing the AI peak is the most important thing for ByteDance right now," and that Volcano Engine's MaaS business "is becoming ByteDance's foundational business." The focus of commercialization AI is shifting from the internal efficiency proposition of "using generative AI to optimize advertising delivery" to the external growth proposition of "selling model capabilities as cloud services." When MaaS and API calls become the center of the narrative, the organizational necessity of an independent GenAI R&D team built around advertising and marketing scenarios naturally dissolves.
Wen Shilei's next stop happens to fall at the other end of this shift.
Octopus Dynamics was founded in January 2026 and focuses on embodied intelligence foundation models and robot products. Its founder, Du Dalong, was formerly a core researcher at Baidu's IDL deep learning lab and an early key technical member of Horizon Robotics. Less than two months after its founding, the company completed a seed round of nearly $50 million, with investors including Horizon Robotics, Hillhouse Venture Capital, Xiaomi Strategic Investment, and Shunwei Capital. Its cumulative financing has reached nearly 1 billion yuan, and its valuation has reached unicorn level.
From the engineering architecture of commercial advertising to the perception-decision-execution loop in the physical world, Wen Shilei's choice echoes, in a certain way, the destinations of the other two leaders of ByteDance's commercialization GenAI team—Google's commercial video foundation models and world model entrepreneurship: the next leg of generative AI is moving from content and delivery on screens to model capabilities and physical interaction beyond screens.
And the advertising AI monetization line that ByteDance left behind has had its mission reassigned to Volcano Engine's MaaS ledger, with the disappearance of the organizational designation.