Smart Prescriptions for Yunnan Baiyao Group's Industry Transformation
Prodotti, soluzioni e servizi per le aziende
As living standards improve and populations age at an increasing rate, the public interest in progress of the pharmaceutical and healthcare sectors has grown significantly. This is especially apparent in China, where traditional Chinese medicine (TCM) companies are being compelled to adopt new production methods and business models by integrating biotechnology (BT) and next-generation information technology (IT). This not only generates economic value, but also lays a solid foundation for improving public health.
Yunnan Baiyao Group, a century-old Chinese brand, is embracing transformation through a partnership with Huawei. The collaboration has enabled the pharmaceutical giant to digitize, visualize, and trace its products across the entire value chain, from medicinal herb cultivation to R&D, production, and service. By combining Yunnan Baiyao Group's extensive expertise and heritage in the TCM industry with Huawei's Pangu models, the two companies have created the Leigong Model specifically tailored for the industry. This partnership unlocks the value of AI by developing an industry-specific model on top of general foundation models, serving as a model for digital transformation in China's TCM industry.
Yunnan Baiyao Group, a renowned pharmaceutical company, owes its name to the proprietary traditional Chinese medicine formula Yunnan Baiyao (Chinese for "white medicine from the Province of Yunnan") developed by practitioner Qu Huanzhang in 1902. After experimenting with various medicinal herbs, Qu came up with this groundbreaking remedy to treat bruises, injuries, and internal bleeding. He started a family business in the 1930s, which later became the state-owned pharmaceutical company that we know of today.
Today, a century after the invention of the medicine, Yunnan Baiyao Group has evolved into a leading national brand with a diverse portfolio of 416 products across 40 categories, encompassing natural medicines, prepared drugs, medical devices, and dietary supplements. Notably, 16 of these products have achieved sales of over CNY1 billion. In 2023, the company secured the top position in China's pharmaceutical industry in terms of brand value.
Yunnan Baiyao Group has undergone a significant transformation from a traditional manufacturer of TCM products to a modern company offering a wide range of health solutions. The company focuses on creating an ecosystem that encompasses disease prevention, health preservation, diagnosis, treatment, and rehabilitation, providing customers with advanced and comprehensive health solutions. The company has proactively adopted AI as a key enabler of this transformation, developing a dedicated AI model for the pharmaceutical industry that allows them to produce with excellence.
The discipline of pharmaceuticals is known for its strict standards. To guarantee the efficacy and safety of drugs, Yunnan Baiyao Group, while staying grounded in TCM theories and practical experience, is dedicated to optimizing the entire value chain through AI to make Chinese medicinal material production traceable from start to finish. This is crucial for developing high-quality Chinese medicines. And by harnessing the power of AI, the company aims to provide the public with accessible TCM information, improve Chinese medicine R&D and marketing, classify medicinal materials, offer guidance on Good Agricultural Practices (GAP) during herb cultivation, and assist in medicine processing and production. Through these smart applications, Yunnan Baiyao Group seeks to enhance people's well-being by tapping into the robust capabilities of large AI models in knowledge retrieval, agile computing, and interactive experiences.
This endeavor entails considerable efforts from Yunnan Baiyao Group. First, they must establish seamless communication between internal supply chain information systems, such as the Enterprise Resource Planning (ERP) system, Manufacturing Execution System (MES), and Warehouse Management System (WMS). To achieve data-driven operations, they also need a sophisticated AI model, integrated data, and enhanced data processing capabilities. For another, the company must incorporate its own industry know-how into the AI model to create a customized technology platform, which provides essential engineering, modeling, and application capabilities tailored to the TCM industry.
For Yunnan Baiyao Group, the key value of large AI models lies in its ability to seamlessly connect the entire industry ecosystem and support in-depth interactions with customers.
Huawei Pangu models are ideal for these needs thanks to Pangu's 5+N+X three-layer decoupled architecture1 , which offers excellent openness. With Pangu's AI computing backbone, AI platform, and development tools, Yunnan Baiyao Group can effectively glean value from data and enhance every aspect of their value chain. This includes activities such as plant breeding using high-quality seed sources, Chinese medicinal material trading and processing, intermediate product production, patent medicine and preparation production, R&D and production of dietary supplements, and sales of finished products.
Notoginseng, also known as Sanqi, is a key ingredient in the Yunnan Baiyao formula, and Yunnan Baiyao Group holds the largest market share for this medicinal herb in China. The company has created a digital platform that covers Sanqi's entire product lifecycle, from seed breeding, planting, and initial processing, to refined processing and extraction, pharmaceutical and food manufacturing, and finished product distribution, allowing for end-to-end traceability and visualization. The platform not only enables the sharing of professional knowledge to downstream sectors, which drives business growth, but also communicates user requirements to upstream enterprises, fostering business innovation.
To maximize the value of Huawei Pangu models, Yunnan Baiyao Group has been consistently updating and integrating their industry know-how into the models. Through this collaboration, the two companies have developed the Leigong Model specifically designed for the medical field with a particular focus on TCM. This model has been trained on high-quality data from over 10 million pieces of TCM literature, and is equipped with prescription and drug knowledge graphs that contain over 1 million entity relationships. It has since supported AI applications in a range of scenarios, including Chinese medicinal material value chain service, TCM decision-making assistance, knowledge popularization service, and digital marketing.
• Chinese medicinal material value chain service: The large Leigong Model, combined with more specialized smaller models, allows Yunnan Baiyao Group to deploy AI applications in diverse scenarios across the entire Chinese medicinal material value chain. These applications include AI-powered identification and sorting of medicinal materials, GAP guidance during herb cultivation, and intelligent planting management.
• TCM decision-making assistance: Leigong is able to gain in-depth insights into industry trends and knowledge, allowing it to make scientific, data-driven decisions. Huawei ModelArts has been instrumental in the model's development, completing tasks like model training through incremental pre-training, data processing for understanding ancient texts and combinations of ancient and modern texts, data proportioning, and data synthesis. This helps improve data quality and prevent models from forgetting what they have learned from previous data. In addition, a prediction model has been built to enhance Leigong's judgment-making capabilities in complex scenarios, especially in diagnosing illness and diseases using TCM principles.
• Knowledge popularization service: Leigong provides users with a variety of services ranging from self-diagnosis assistance to health management, offering comprehensive information and explanations on TCM, including formula ingredients and symptom interpretations. All of this knowledge is traceable and cited from credible sources. This approach significantly reduces model hallucinations, minimizing misleading outputs from AI models.
• Digital marketing: Leigong enables seamless collaboration between different sections along the value chain by offering digital communication, negotiation, and electronic transaction services. By integrating online and offline marketing channels, the platform boosts business growth and elevates the overall user experience for both partners and users.
The Leigong Model is named after a renowned ancient Chinese physician, Lei Gong, who was a minister of Huangdi (Chinese for the "Yellow Emperor") and a skilled doctor with extensive knowledge on drug processing and preparation. In a dialogue on medicine with Huangdi, Lei Gong pointed out the importance of not only clearly understanding principles and prescriptions explained in medical literature but also applying this knowledge to clinical practices.
Today, in the age of AI, Yunnan Baiyao Group continues the tradition of combining medical knowledge and practice through the Leigong Model. The model safeguards people's health and well-being, showcasing the company's commitment to integrating its century-old medical expertise with Huawei's cutting-edge AI technologies. This has empowered the pharmaceutical company to develop multi-modal solution tailored to the TCM industry and create an exemplary industry-specific large AI model.
Yunnan Baiyao Group and Huawei have also agreed to establish an alliance to further develop AI models for the TCM industry. They envision a robust and open industry ecosystem that harnesses the accumulated wisdom of TCM using digital and intelligent technologies, as well as enabling efficient collaboration and sharing. Through technological innovation, the two companies are set to tap into the full potential of TCM and unlock infinite possibilities for this longstanding industry.
1 The architecture of Pangu models consists of layers L0, L1, and L2. L0 consists of five foundation models, L1 contains numerous industry-specific models, and L2 provides even larger numbers of models for more specific scenarios and tasks.
