Three Rounds of Financing Secured in Half a Year, This Company Enters the Children's Growth Interaction Track with AI and Data | 36Kr Exclusive

Source / Enterprise

This article is about 3000 words, with an estimated reading time of 7 minutes

Author | OU Xue

Editor | YUAN Silai

Hard Krypton has learned that PeakRidge Digital Technology (hereinafter referred to as "PeakRidge Technology") has recently completed tens of millions of yuan in Angel+ round financing, led by Horida Investment, with participation from Suzhou Yida Fund and Hexa Capital. The funds will be mainly used for the prototype R&D of the first-generation educational robot, expansion of the technical team and construction of the sales system.

Founded in October 2025, PeakRidge is a digital technology company focused on data assets of children's growth. Its founder Jiang Xiaodi graduated from Cornell University in the United States, and previously founded Xingluan Tianxia, a leading knowledge payment and MCN company, with years of experience accumulated in the fields of educational traffic operation and educational product R&D. The co-founders and partners of the team have backgrounds covering the embodied intelligence industry, AI chip R&D, top IPs in the education sector and leading listed companies in high-end manufacturing.

At present, PeakRidge Technology's product system is divided into two lines: software and hardware.

On the software side, Magic Class APP, an AI learning tool focused on home learning scenarios, has launched content subscriptions and services for mathematical thinking and English learning, and will subsequently launch courses of Chinese literacy, AI general knowledge and robot programming in 2026. Different from most AI course products on the market, Magic Class APP does not merely digitize course content. While emphasizing fun AI interactions, it has built an AI data middle platform at the underlying layer. Through granular data collection, including the stay duration of each animation frame, repeated error rate of each question, exit nodes during use, and personalized feedback of users in interactions and other usage behaviors, it generates multi-dimensional learning reports for each child.

Introduction to English courses on Magic Class APP (Source / Enterprise)

"We have developed our own small model based on the open-source model, with adjustments made for children's language habits and learning content," said Jiang Xiaodi. The APP currently integrates functions such as AI oral English foreign teacher, wrong question collection, and phased learning suggestions. Meanwhile, mini-programs for parents and teachers have been developed to synchronize learning progress and reports in real time.

The hardware side is the core focus of PeakRidge Technology in the next stage. The desktop companion learning robot under development by the company adopts a wheel-legged solution instead of a bipedal solution, and the first-generation prototype is expected to be completed before November this year.

The reason Jiang Xiaodi gave for choosing wheel-legged instead of bipedal is quite pragmatic: "Aside from cost and technology, the biggest problem of bipedal robots is that they often have the possibility of falling. The wheel-legged structure is more stable and has higher safety attributes for children in the home environment."

Schematic diagram of the desktop companion learning robot (Source / Enterprise)

It is worth noting that the functional positioning of this robot targets the mid-to-high-end learning machine market, rather than the two common types of educational robots on the market -- those with toy attributes and those with programming education attributes.

Hard Krypton further learned that the product is priced at around 6,000 yuan, with voice interaction as the main function. At the same time, the built-in model in the camera monitors the child's sitting posture, micro-expression, attention concentration and emotional feedback. In addition, the robot itself does not have a screen. When it is necessary to display text content, it is completed through screen projection or linkage with the Magic Class APP.

"Many children have a problem when using tablets, that they will use non-educational functions to do things unrelated to learning," Jiang Xiaodi said. "What we provide is actually a customized tutor. We use robots to replace tablet learning machines, and add functions of behavior detection, emotional companionship and health detection."

Regarding the competitive landscape, Jiang Xiaodi's judgment is that there are no direct competing products on the market at present. In his view, the existing educational robots are either of toy attributes with low technical threshold but low unit price, or programming educational robots targeting a non-just-needed market and a single programming literacy subject. PeakRidge Technology's companion learning robot cuts into the home scenarios and learning tracks that every child and parent needs.

"We are targeting the most just-needed and largest market. The demand for after-school all-subject education and family companionship that every middle-class and above family has to face," Jiang Xiaodi said.

He further explained the logic of hardware profitability: "The domestic hardware manufacturing industry is already very mature. Thanks to our background in the industrial chain of listed companies, the BOM cost per unit can be controlled at a level far below the average market cost. Moreover, different from other pure hardware products, our robot has many built-in long-term service and companionship functions. Through back-end subscription services and function renewals, the business model will be more long-term and healthy."

If it only achieves this step, PeakRidge is just a "hardware plus content" company. The real difference lies in its layout logic of data assets.

According to PeakRidge's plan, the product line will not only stay at robots and APPs. It will subsequently deploy smart hardware such as children's health bracelets and sports hardware to cover data collection in multiple dimensions including learning, behavior, health and social interaction.

After these data are imported into the background, a vertical growth database will be generated for each child, with a time span of 6 to 12 years.

At present, the total number of users of Magic Class APP is close to 300,000, with a broad user base. After the mass production of the companion learning robot, the C-end shipment volume in the first quarter is expected to exceed 10,000 units.

Jiang Xiaodi explained the rationality of this goal from two dimensions: "Compared with the shipment volume of the learning machine market, 10,000 units is not a large number in the just-needed education scenario. If placed in the entire embodied intelligence market, few companies may be able to achieve it, but when it comes to the education scenario, this is a conservative number instead."

In addition, the company's commercialization path next year is divided into two channels: C-end direct sales and B2B2G school entry. The software side has opened free trials to some schools to obtain precise traffic, while the hardware side strives for class procurement orders by cooperating with schools to customize functions.

The following is an edited excerpt of the conversation between Hard Krypton and founder Jiang Xiaodi:

Hard Krypton: What is the essential difference between your companion learning robot and the learning machines and educational robots on the market?

Jiang Xiaodi: Why are parents willing to spend thousands of yuan on learning machines? Because the ability to improve learning performance is a rigid demand. Our robot first has all the educational functions of a learning machine, but on this basis, it adds the capabilities of behavioral education and psychological education, including behavior monitoring, emotional support and emotional companionship. At the same time, it can liberate parents, so that parents no longer need to invest a lot of time and energy to accompany their children in dictation, check homework, answer questions, and pay attention to whether the child's eyes are too close to the textbook, etc. This is not a cold tablet, but a customized tutor and children's companion that every family can afford and urgently needs.

Moreover, one of the most troublesome problems for parents is that their children secretly play games with tablets. Our robot does not have this problem, and the child's sitting posture, concentration degree and emotional changes during the learning process can be fed back in real time. This information is also very important for parents and teachers to understand the child.

Hard Krypton: How to ensure compliance when doing data business in children's scenarios?

Jiang Xiaodi: Our data collection must be authorized by guardians, and the authorization is itemized and revocable. Data desensitization is the foundation. When children's data enter the cloud model, specific identity information will not be associated. Although the robot is equipped with a camera, behavior perception is calculated locally, and only structured events and desensitized features are uploaded to the cloud. The original images will never leave the home. We have also designed a deletable and migratable mechanism. After users terminate the service, they can erase all vertical archives with one click. Compliance is not only about avoiding risks, transparent data management itself is the trust selling point of the product.

Hard Krypton: What is your competitive barrier compared with the hardware products of educational institutions?

Jiang Xiaodi: The core competence of traditional education companies is to produce educational content on the assembly line and replicate teaching methods. But what we are doing is essentially combining AI technology, embodied robots and mature educational content and methods to meet the multi-dimensional application needs of children's learning and home scenarios. At the same time, on the hardware side, we have advantages that educational institutions with educational content and teaching methods as the core do not have, in terms of supply chain cost, technology R&D and iteration speed. In addition, with the accumulated MCN and influencer resources on the sales side, our market expansion speed can be faster.

Investor Views:

He Junli, General Manager of Horida Investment: The industrial development of artificial intelligence and embodied intelligence is entering the deep-water zone. Whether relying on the general large model base to serve C-end users in specific vertical tracks and commercial scenarios is the core leap for startups to complete the transition from technical prototypes to commercial scenarios, and it is also an important node to truly verify whether technical capabilities can serve the market. PeakRidge enters campus and home scenarios, focuses on the high-frequency rigid-demand vertical scenarios and crowds of parents and children, and quickly completes product polishing and scenario closed-loop verification, which is expected to seize the first-mover advantage in the early window of industry explosion.

Zhang Chenjian, Managing Partner of Suzhou Yida Fund: The behavior data of young people has long had the industry pain points of high value density and high difficulty in structuring. PeakRidge has built a positive flywheel of "scenario implementation, data precipitation, and model iteration". Through real human-computer interaction, it continuously precipitates exclusive vertical datasets for the AI Generation crowd in multiple dimensions such as academics, behavior, health and social interaction, feeds back the iteration of vertical models, and gradually builds a long-term moat that cannot be replicated quickly. It has even seized the earliest core data assets of the next generation of people in the most valuable "data" track in the AI era.

Rong Bo, Managing Partner of Hexa Capital: The core team of PeakRidge has both solid industrial accumulation and professional technical background: on the one hand, it has senior professional R&D background in the field of embodied intelligence and Agent model R&D, and at the same time has mature practical experience in commercialization of education industry, home and campus channels and parent services. On the other hand, it deeply covers core technical links such as algorithm R&D, data collection and analysis, and hardware engineering. The team has the capabilities of underlying technology breakthrough, industrial implementation and sustainable commercial realization at the same time, which is a very rare compound entrepreneurial team in the current embodied intelligence and AI track.