Kidodo is an AI-powered learning companion robot designed for children aged 0–10, aimed at becoming a child’s “smart growth partner.” With multimodal interaction as its core design philosophy, Kidodo can “see, hear, and respond” to children, helping them learn languages, develop reading habits, and expand their cognitive boundaries through natural conversation. Its design features soft, rounded shapes made with food-grade silicone and an interchangeable "wearable system," creating a warm, friendly, and lifelike emotional experience. Kidodo uses an end-to-end real-time multimodal model, enabling high-precision visual recognition and natural speech interaction. It supports 23 languages, including English, Chinese, Japanese, and Spanish, and can recognize and explain picture books, magazines, doodles, and real-world objects. With an innovative private-first architecture powered by PrivateLoRA technology, Kidodo ensures data security and privacy. It also features recyclable materials and modular construction, promoting sustainability and extending the product's lifecycle. Kidodo blends cutting-edge AI with emotional design to create a truly unique and safe experience for children and parents alike.

This project addresses shared global challenges in early childhood education through in-depth exploration of functional innovation, emotional connection, and user experience. Families around the world face similar constraints: in China, over 70% of children are primarily cared for by grandparents, resulting in generational gaps in language and cognitive stimulation; in South Korea, early education and English kindergarten tuition can reach prohibitively high monthly costs; in Japan, childcare and household responsibilities disproportionately burden mothers, leading to severe time constraints; in Europe, multilingual learning is common, yet family support resources are fragmented; and in North America, parents place strong emphasis on reading education, while declining陪读 time and high private tutoring costs limit sustained involvement. In response, Kidodo was created as an AI family companion that can accompany, interact with, and understand children’s developmental needs. Powered by AI empathy and real-time multimodal cognition, Kidodo offers screen-free immersive interaction, continuous learning support, and emotionally responsive engagement. By easing parental time pressure while providing children with consistent, high-quality educational companionship, the project enables families to remain meaningfully involved in their children’s growth, even under limited time and resource conditions.

The project demonstrated strong commercial performance following its launch. Within the first week of release, Kidodo achieved over 10,000 units sold, enabling rapid market penetration in the AI-powered early childhood education and learning companion segment, and validating the product’s positioning and technological approach. In terms of brand and communication impact, the project established a clear and differentiated market identity centered on “AI empathy and multimodal cognition,” successfully extending the brand from a technology-driven background into the family-oriented intelligent hardware space. As the company’s first hardware product, Kidodo built an early user base while significantly enhancing brand recognition, credibility, and long-term commercial potential in the children’s education and family AI market.

Kidodo redefines early childhood education through the integration of AI empathy and multimodal cognition. By transforming traditional screen-based, passive learning into immersive, voice-driven interaction, the project advances the industry toward more human-centered, emotionally engaging, and personalized educational experiences. As one of the first products to deeply integrate an end-to-end multimodal large model into real-world children’s learning scenarios, Kidodo establishes a complete interaction loop encompassing vision, voice, and emotional feedback—elevating AI from a functional tool to an empathetic learning and companionship partner. Culturally and educationally, its cross-language and cross-cultural adaptability enables flexible deployment across diverse families and education systems, contributing to more inclusive access to quality early education. From a sustainability perspective, Kidodo’s localized privacy computing and “privacy-first” architecture set a new benchmark for responsible family AI products, promoting a safer, more ethical, and people-oriented direction for the intelligent hardware industry.

Core Team: - Dr. Zeng Xiaodong, Founder Dr. Zeng, a disciple of internationally renowned computer scientist Isabel Tracoso, is one of China’s early pioneers in machine translation. He was selected for the MIT TR35, a list of the top 35 innovators under 35, which also includes notable AI entrepreneurs such as Kimi Yang Zhilin, Meituan’s founder, and Kuaishou’s Su Hua. With over 10 years at Alibaba and Ant Group, Dr. Zeng led the core engine of Alibaba's machine translation system. He is also a co-founder of Ant Group's Tech Lab, where he spearheaded innovative products like “unmanned supermarkets,” “facial recognition payments,” and “IoT hardware.” - Yang Dongyun, CPO A graduate of the University of Edinburgh, home to 20 Nobel laureates, Yang has won prestigious design awards such as the Red Dot and IF Design Awards. With extensive experience at Huawei, Alibaba, and ByteDance, she has held roles as product manager, design director, and business lead. Yang has led cross-functional teams, including product, user research, design, and growth. Her projects have won accolades such as the Harvard Business Review China New Growth Digital Innovation and Global Golden Trend Empowerment Innovation awards. She is also an active speaker at conferences like PCon Global Product Innovation and Bund Summit and holds over 20 patents. - Alex, Chief Architect A serial entrepreneur and multiple national champion in AI algorithm competitions, Alex holds a Master’s degree from Tongji University and the Hong Kong Polytechnic University. He specializes in building AI-native data intelligence systems using big data, NLP, and knowledge graphs. - Vincent, Model Scientist A Master's graduate from Fudan University, Vincent has published papers at top-tier conferences like ICML and KDD. His research expertise lies in machine learning and data modeling. - Jay, Product Director Jay holds dual degrees in Information Science and Architecture from Cornell University, leading projects that have won numerous international awards. He has successfully incubated new businesses from 0 to 1, expanding them into global markets. - Joe, Business Director With over 10 years of experience in marketing and business management, Joe has won multiple "Outstanding Management" and "Best Team" awards. He has managed PR and go-to-market (GTM) strategies for tech companies listed on the A-shares and US stock exchanges. Joe also has rich experience in smart healthcare and IoT industries.

1.A parallel collaboration workflow integrating design, AI engineering, and educational science was established from the earliest stages. Children’s attention and emotional experience were defined as core metrics, directly informing model architecture and interaction logic, with behavioral testing and emotional evaluation conducted in every iteration to ensure real-world child-centered design. 2. An end-to-end real-time multimodal large model was tightly coupled with children’s educational scenarios. By jointly collecting and synchronously labeling voice, vision, and emotional signals, a closed-loop interaction dataset was formed, enabling 350ms understanding and generation, while 76B-parameter reasoning supports open-ended questioning and real-time knowledge exploration. 3.An emotion-engineering approach was introduced, translating children’s emotional states into computable parameters mapped to 100 eye expressions and 48 emotional responses. Language, tone, and expression are coordinated in real time, with continuous feedback enabling an evolving and sustainable empathetic companionship. 4. A privacy-first AI management workflow was established through PrivateLoRA, enabling localized training and inference. Personalized memory and learning preferences remain on-device, with no sensitive data uploaded to the cloud, setting a scalable and sustainable privacy standard for child-focused AI products.

The project was initiated in December 2024 and progressed through concept development, technical validation, product design, and system integration, leading to its first market launch in September 2025. The overall development cycle spanned approximately 10 months. During this period, the team invested several million RMB in research and design, including in-house multimodal model training, interaction data analysis, emotional system design, child-centered usability testing, and privacy-first AI architecture development. This sustained, cross-disciplinary investment ensured a high level of maturity in both technological performance and emotional experience.

1. Initial Vision and Outcome Evaluation The project was initiated in response to the lack of quality family companionship and the uneven distribution of early childhood education resources. Its core objective was to explore a new form of child-centered learning and companionship that integrates advanced AI technology with emotional care. Beyond technological leadership, the project aimed to create an AI companion capable of genuinely understanding and supporting children’s growth. The final results exceeded expectations, particularly in real-time multimodal interaction, empathetic engagement, and long-term user retention among children. Positive market and user feedback after launch validated the feasibility and forward-looking value of the “AI empathy + multimodal cognition” approach in early childhood education. 2. Strategic Alignment and Business Impact This project is a key implementation of the company’s long-term strategy, AI for Growth, which focuses on building family-oriented AI products grounded in emotional intelligence and privacy-first principles. As the company’s first intelligent hardware product, Kidodo successfully translated this strategic vision into a tangible offering and achieved strong commercial validation. Within the first week of launch, sales exceeded 10,000 units, establishing a leading position in the AI-powered early education segment and laying a solid foundation for the company’s future product ecosystem and brand development.