36Kr Exclusive | The pet health large model company has completed two consecutive rounds of financing, with a layout integrating hardware and software, and has served over 200 pet hospitals.

Autor | Qiao Yujie

Redakteur | Yuan Silai

Hard Kr has learned that Chongqing Qisuanfa Technology Co., Ltd. (hereinafter referred to as "Qisuanfa"), a health company focusing on large models for pets and an ecological enterprise of Zhipu's "Z Plan", recently completed a financing of tens of millions of yuan. The investors are Qifu Capital and Juheng Venture Capital. The funds from this round will be mainly used for product iteration, deepening of model capabilities and market expansion.

Qisuanfa was founded in July 2022. It is a large - model technology company focusing on the field of pet health. Relying on the capabilities of multi - modal large models, the company has created a pet health brand that integrates online and offline and coordinates software and hardware.

Chen Li, the founder of Qisuanfa, graduated from King's College London. He is a researcher at Qlalgorithm Lab and also a serial entrepreneur. Liu Yudong, the technical partner, is a doctor from the University of Pennsylvania and has long been engaged in research in the field of AI medical care. Another technical partner, Deng Zihao, graduated from the University of Pennsylvania and focuses on the new - generation edge computing technology.

Previously, with the support of Zhipu's professional model capabilities and the business resources of "Xiaonuan Doctor", the company has built a set of vertical and implementable AI model systems around the pet medical scenario.

In the field of pet medical care, due to the lack of systematic evidence - based data and the inability of pets to actively express their symptoms, the diagnosis is quite difficult. Chen Li introduced that Qisuanfa conducts model training based on tens of millions of pet medical records, medical images and behavior data. Compared with general large models, it has a better understanding of the individual differences in breeds, symptom expressions and diagnosis and treatment logic in pet medical care. The model can not only output diagnosis results, but also provide diagnosis basis, risk warnings, decision - making paths and solutions, which is closer to the needs of real medical scenarios.

At the same time, the company adopts a software - hardware integrated solution of "cloud - based large model + edge - end NPU deployment", enabling AI capabilities to truly enter pet hospitals, pet smart wearable devices and home scenarios, rather than remaining at the level of simple Q&A.

In terms of business promotion, the company has completed the application for a number of patents related to large models, filings and Internet hospital licenses, and has made phased progress in terms of compliance and professional qualifications.

Currently, Qisuanfa's auxiliary consultation system has realized the function of assisting doctors in receiving patients and is open for free use by pet doctors. Chen Li introduced that the platform has served more than one million times in total, cooperated with more than 200 hospitals, about 3,000 doctors have registered and used it, and the daily active users of the platform are close to 5,000. A relatively stable closed - loop of data feedback and model training has been formed.

Specifically, doctors can use the AI consultation model to improve the efficiency of daily patient reception, and the data generated during the use process will continuously feed back to the model training. At the same time, the company's cooperation with pet hospitals not only stays at the level of system access, but also extends to the recommendation of diagnosis and treatment services. After users complete the consultation, the platform can provide users with subsequent services such as drug recommendations and hospital referrals based on the capabilities of the Internet hospital license, thus forming a complete link of "consultation - diagnosis and treatment - medication - data feedback".

In terms of hardware, the company launched the Pachi Pet AI smart collar in the early stage, and it has now been iterated to version 3.0. Compared with the previous solutions, the new - generation product can run on the full edge side without additional host devices, and has been optimized in terms of edge computing power consumption control and algorithm stability.

This collar weighs only 19 grams, is mainly targeted at cats and dogs, and is "the world's lightest pet smart wearable device". The cumulative sales volume is close to 20,000 units. The product can automatically complete functions such as pet positioning, real - time posture prediction, status recognition, movement, eating and sleep tracking. Users can obtain basic health data without additional operations.

(Quelle: Unternehmen)

In addition to the smart collar, the company's AI feeder has completed the pre - sale of 1,000 units, realizing exclusive scientific feeding. Products such as the AI ICU, which features AI automatic early warning and monitoring, have also begun to be gradually implemented in more than 30 pet hospital scenarios.

(Quelle: Unternehmen)

In addition, in the emotional companionship scenario, the company has also reached a cooperation with OPPO. Taking OPPO's official theme store as an example, its AI pet desktop wallpaper function uses the relevant interface capabilities provided by Qisuanfa, covering more than hundreds of thousands of users.

Next, Qisuanfa plans to further build a Q&A search and recommendation engine for the pet industry, hoping to become an infrastructure platform in the field of pet health management and provide exclusive health management and services for every furry friend.

Following are the selected excerpts from the interview (slightly edited):

Hard Kr: There are a wide variety of pet breeds. Will the "thousands of pets with thousands of faces" make it difficult to establish the generalization ability of the model?

Chen Li: We currently have two main solutions. First, as the user scale continues to grow, our basic model will become more and more accurate. Because the coverage of data on pet behavior, breeds and living environments will continue to increase, the generalization ability of the model will also continue to be enhanced.

We support users to conduct personalized AI training. For example, if there are differences between the behavior habits of some pets and the standard data, users only need to upload a video for feedback, and we can quickly complete targeted behavior learning and generate an exclusive model for the user within a few minutes.

Hard Kr: What are the core differential advantages of Qisuanfa in the pet smart hardware industry?

Chen Li: First of all, it is the team background and technical accumulation. Our core team has long been engaged in research in the fields of AI medical care, multi - modal models and edge computing, which is relatively rare in the pet AI track. The company does not simply integrate solutions, but realizes independent research and development from the underlying algorithms, framework optimization to hardware design. At present, except for the chips, the software and hardware, algorithms, frameworks and edge - side deployment capabilities are all self - developed by the team.

More importantly, in the field of pet models, which has a relatively low cost - effectiveness in terms of input and output, we have opened up a complete closed - loop from behavior data collection, analysis and interpretation, health report generation, to drug recommendation, hospital connection and medical data feedback. A high technical barrier has been established, which is difficult for new start - up companies to replicate.

Many companies in the industry may only stay at the single - point function level. We hope to build a truly sustainable pet health infrastructure.

Hard Kr: What are the company's plans for the next step?

Chen Li: In the future, we will focus on building a Q&A search and recommendation engine for the pet industry, hoping to become the core entrance and infrastructure in the field of pet health management. Currently, the pet nutrition industry still lacks evidence - based medical support, and many product formulations rely more on experience. We hope to gradually establish a more scientific and data - driven pet health recommendation system through long - term data accumulation. For example, by analyzing behavior data such as pet eating duration, frequency and activity status, we can provide users with more accurate health and nutrition advice.