

Leiphone · Jingxi has learned exclusively that former Honor senior algorithm engineer Xiao Yuwei has left the company to start a business, founding AI interaction hardware company KnowVoice in the middle of this year. The company is currently being incubated at the Shenzhen Institute of Science and Technology Innovation under Professor Li Zexiang, and has received early investment support.
Founder and CEO Xiao Yuwei was born in 1998 and graduated with a bachelor's degree from Guangdong University of Technology. He is a co-founder of the GDUT Robomaster team DynamicX. After graduation, he worked at Honor, with a career spanning underlying technology and strategic business.
Xiao Yuwei initially served as a GPU-Turbo-X SE systems engineer, responsible for rendering technology and on-device high-performance optimization. The AI render-display separation project he led, "Honor Phantom Engine," mainly addressed issues of occasional frame drops and sudden stuttering in games. In addition, he holds dozens of domestic and international patents and has represented Honor in standards-setting for the graphics industry.
Later, Xiao Yuwei was transferred to the President's Office as a researcher, shifting toward gaming and AI terminal devices. He led research on Gen Z gamers, terminal gaming experience, and the future development of AI terminals, reporting directly to CEO Li Jian.
It is understood that the project's core team has been assembled, with members from companies such as Huawei, Tencent, and OPPO as well as universities.
Product and technology lead Li Jiawei is a former senior RF engineer at OPPO, with experience in product interaction design and mass production. He led the smartphone industry's first "buttonless AI intelligent edge interaction." Embedded software lead Lin Shenghui has worked at Huawei, Mean Well, and other companies, responsible for embedded project development. Algorithm lead Dr. Li Hongzhe graduated with a bachelor's degree from Harbin Institute of Technology and a doctorate from Korea University. His research focuses on AI software and system security, context management, and memory management, and he previously served as a senior algorithm engineer at Huawei and Tencent.
The first product launched by the KnowVoice team is an on-device input device in pendant form, focused on understanding personal intent and assisting expression.
According to the introduction, the product does not replace the original input methods of devices such as phones and computers, but instead integrates into different applications in a floating form: whether in WeChat chats or document writing, users can invoke information previously accumulated by the device to help supplement content.
This is also a different approach from that of current mainstream AI memory hardware.
At present, KnowVoice already has a product prototype. The team is using the Vibe Coding community as its early entry point, mainly targeting professional users who need high-frequency expression and information processing, such as product developers, hardware engineers, structural engineers, self-media bloggers, and independent OPC entrepreneurs.
In terms of data processing, the team has chosen to keep core data on-device, not upload it to the cloud, and not route it through company servers, hoping to return data control to users.
Leiphone (official account: Leiphone) · Jingxi will continue to follow the team's subsequent progress. Welcome to add the author's WeChat frenchfries0016 for communication.
This is an original article by Leiphone. Reproduction without authorization is prohibited. For details, see the reprint notice.