The test results cannot show which part the child is stuck on, but eye movement data can: Eyemind combines visual perception technology and large educational models to build a fully personalized AI home tutor that offers customized experiences for every individual user.
Recently, Beijing Shizhu Technology Co., Ltd. announced that its Eyemind AI Tutor project has launched a new round of financing. Relying on multimodal perception and adaptive learning technologies, the project targets core pain points in K12 family education scenarios, including the lack of personalized tutoring and the difficulty in quantitatively tracking learning status, and strives to provide Chinese families with smart education solutions that balance high accuracy and wide accessibility.
Just-in-demand pain points spawn a hundred-billion-level market, with significant mismatch between supply and demand in the traditional tutoring model
Education has always been a rigid expenditure for Chinese families. According to industry research data, education expenditure accounts for 36.5% of the total expenditure of Chinese families, and as high as 78% of the total expenditure on children. The number of K12 students across the country exceeds 190 million, and the structural contradiction of uneven distribution of high-quality educational resources has long existed — the gap in teacher resources between first-tier cities and county-level regions can be more than 100 times.
Against the background of the continuous deepening of the "double reduction" policy, off-campus discipline training has been strictly regulated, and family education scenarios are facing multiple practical dilemmas: offline one-on-one tutoring is expensive and has fixed schedules, which is difficult to adapt to students' fragmented learning needs; traditional learning machines and teaching aid tools have single functions, mostly limited to question bank search and standard answer output, and cannot realize real personalized teaching; parents generally lack professional educational capabilities, making it difficult to accurately identify their children's weak knowledge points and changes in learning status, and tutoring homework has become the focus of contradictions in many families.
At the same time, the maturity of AI technology provides basic conditions for industry reform. In April 2026, five departments including the Ministry of Education jointly issued the "Artificial Intelligence + Education Action Plan", which clearly put forward the goal of basically forming a pattern of deep integration of artificial intelligence and education by 2030. Driven by favorable policies, the AI smart education track has entered a hundred-billion-level scale. In 2025, the market size of GenAI + education reached 344.2 billion yuan, with an annual compound growth rate of 37%. It is expected to rise to 891 billion yuan by 2028, among which the C-end family education scenario has become one of the fastest-growing segmented fields.
Multimodal perception builds a closed teaching loop, and eye tracking technology empowers accurate learning situation diagnosis
The core differentiation of the Eyemind AI Tutor project lies in the deep integration of computer vision, eye tracking technology and large educational models, forming a complete closed loop of "perception - diagnosis - teaching - feedback". Different from most products on the market that only judge the mastery of knowledge through test results, this project attempts to use non-intrusive visual perception technology to capture students' behavior data such as eye movement trajectories, gaze duration, and micro-expression changes in real time, and combine the research results of cognitive psychology to quantitatively evaluate students' attention concentration, cognitive load level and understanding of knowledge points.
Academic research shows that there is a significant correlation between students' eye movement data and learning status. When students have difficulty understanding knowledge points, they usually show characteristics such as scattered fixation points, increased number of regressions, and changes in pupil diameter. Based on this principle, the Eyemind system can identify students' cognitive bottlenecks in real time during the learning process, and dynamically adjust the teaching rhythm and explanation strategy — it actively introduces interactive links when detecting that students are distracted, and automatically switches explanation perspectives and supplements example questions when judging that knowledge points are not mastered, truly realizing adaptive teaching tailored for every unique student.
In terms of content, the system covers multiple versions of teaching materials for all stages from primary school to high school, and builds a structured subject knowledge graph, supporting functions such as intelligent homework tutoring, error cause analysis, weak knowledge point induction, and oral correction for Chinese and English. At the same time, a complete teaching safety mechanism is set up to avoid directly giving answers and weakening students' thinking ability, and cultivate students' independent problem-solving thinking through guided questioning and step-by-step decomposition. The parent side can view the learning report in real time, intuitively understand the child's learning progress, knowledge mastery status and concentration changes, and solve the problem of information asymmetry in the process of home-school communication.
Lightweight service model lowers the threshold for use, and technology implementation still requires market verification
In terms of business model, the Shizhu Technology team plans to adopt a lightweight operation model of "hardware + service subscription". Different from the one-time investment of thousands of yuan for traditional learning machines, the project intends to launch software services adapted to ordinary tablets and computer devices to lower the access threshold for family users, and also provide supporting smart hardware terminals for users to choose from. The revenue sources include monthly/annual membership subscriptions, value-added content packages, hardware sales, etc., with the unit customer price positioned at an affordable range for general public families, targeting the inclusive education demand in the sinking county-level market.
Industrial and commercial information shows that Beijing Shizhu Technology Co., Ltd. was established in May 2026, with its registered address at Zhongguancun Street, Haidian District, Beijing. Its legal representative is Wang Zixin, and its business scope covers fields such as basic artificial intelligence software development, artificial intelligence theory and algorithm software development, and smart hardware sales. The core members of the team come from cross-cutting fields such as artificial intelligence, educational technology and cognitive psychology, and have experience in multimodal interaction technology R&D and education productization.
It is understood that the current Eyemind AI Tutor product has completed core algorithm verification and prototype development, and is continuously iterating and optimizing in a small-scale user test stage. The new round of financing funds will be mainly used for multimodal algorithm iteration, educational content ecology construction, product polishing and market promotion.
Industry analysis points out that the current AI education track has entered a critical period of large-scale application from the technology pilot stage, but it also faces many challenges: whether the accuracy of perception technologies such as eye tracking on ordinary devices is sufficient to support accurate teaching judgment, whether AI interaction can truly replace the emotional support and value guidance of human teachers, and how to ensure data privacy and minor protection, all of which require the entrepreneurial team to continue exploring in the process of technology implementation. Whether the Eyemind project can occupy a place in the fierce market competition with its differentiated technical path remains to be tested by the market performance after the official launch of the product.